Wikipedia:Village pump (proposals)

 Policy Technical Proposals Idea lab WMF Miscellaneous 

• Table of contents
• First discussion
• Jump to bottom
• New topic
• All village pumps

"WP:PROPOSE" redirects here. For proposing article deletion, see Wikipedia:Proposed deletion and Wikipedia:Deletion process.

The proposals section of the village pump is used to offer specific changes for discussion. Before submitting:

• Check to see whether your proposal is already described at Perennial proposals. You may also wish to search the FAQ.
• This page is for concrete, actionable proposals. Consider developing earlier-stage proposals at Village pump (idea lab).
• This is a high-visibility page intended for proposals with significant impact. Proposals that affect only a single page or small group of pages should be held at a corresponding talk page.
• Proposed policy changes belong at Village pump (policy).
• Proposed WikiProjects or task forces may be submitted at Wikipedia:WikiProject Council/Proposals.
• Proposed new articles belong at Wikipedia:Requested articles.
• Discussions or proposals which warrant the attention or involvement of the Wikimedia Foundation belong at Village pump (WMF).
• Software changes which have consensus should be filed at Phabricator.

Discussions are automatically archived after remaining inactive for 7 days.

« Archives, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232

RfC regarding German nobiliary titles

Should German nobiliary titles inside names be clarified with the footnote text provided by {{German title}} or with the template User:Joe vom Titan/German title hatnote or another way? Feel free to suggest changes to either template, regarding the text or the template parameters. If it is to be a footnote, there's also the question whether to place it right after the title, after the name or at the end of the sentence. Joe vom Titan (talk) 18:38, 24 August 2026 (UTC)

This RfC follows a discussion at Wikipedia talk:Manual of Style/Biography#Footnote in the middle of bold name?. Joe vom Titan (talk) 18:44, 24 August 2026 (UTC)

The footnotes vs hatnotes question has come up when discussing the family name clarification template: Wikipedia:Village_pump_(proposals)/Archive_188#Method_of_surname_clarification. —Joe vom Titan (talk) 19:14, 24 August 2026 (UTC)

Special:WhatLinksHere/Template:German_title shows the 500 pages currently using {{German title}}. Joe vom Titan (talk) 19:32, 24 August 2026 (UTC)

Survey (German nobiliary titles)

Some time ago I had exactly same problem with Dutch surnames. In one bio I saw the name that contained 'bij' or something like that. I remember I met a strong resistance against my request for clarification. After some time I learned a magic mystery word tussenvoegsel and started using it in {{surname}} pages, such as Ten Bos (which is not the same as 10 Bos :-) --Altenmann >talk 19:05, 24 August 2026 (UTC)

As proposer, I'm all for the new hatnote. It has several of advantages. Footnote placement is awkward: The footnote would belong right after the title but that would cut up the name which inhibits the reading flow. The new hatnote is brief and focuses on what's important. It also has the in1919= parameter to distinguish the legal situation before 1919 (which doesn't require explanation) from those after 1919 in Austria and Germany (with the corresponding explanations). Joe vom Titan (talk) 19:11, 24 August 2026 (UTC)

• Support footnote. Some texts are waaaay to long for a hatnote. --Altenmann >talk 19:47, 24 August 2026 (UTC)

  The longest possible hatnote text with the new template is the following: {{User:Joe vom Titan/German title hatnote|Reichsfreiherr|Herzog|in1919=de}}, rendered as

This name includes the German nobiliary titles Reichsfreiherr, translated as 'baron of the empire', and Herzog, translated as 'duke'. In 1919 nobiliary titles and particles (von, zu, etc.) became part of the surname in Germany.

That's not crazy long. Joe vom Titan (talk) 20:09, 24 August 2026 (UTC)

Not crazy, but longish. And it is not tnat vital for understanding so as to stick it on top. I also do not see why the footnote mark must sit right on the title; it may well be after the whole surname: the footnote starts with "This name includes...", i.e., technically the footnote is about name not about title.--Altenmann >talk 20:18, 24 August 2026 (UTC)

Fortunately, few if any articles will use the longest possible example and many contain just one name to be addressed this way. A longer-than-usual hatnote would be an improvement over the the current practice in an article like Karl Ernst von Baer, which opens with: Karl Ernst Ritter von Baer Edler von Huthorn (Russian: Карл Макси́мович Бэр; 28 February [O.S. 17 February] 1792 – 28 November [O.S. 16 November] 1876) was a ... That's an awful lot to get through before finding out what sort of notable person he is and it yields two nearly identical footnotes in the article. —Myceteae‍🍄‍🟫 (talk) 23:51, 25 August 2026 (UTC)

Crazylong. I support a hatnote, but there's absolutely no need for anything like this much detail. Let the link do the heavy lifting, and craft a tight wording for each of the main cases separately: died before 1919, alive in 1919, born after 1919. Only the second needs any sort of mention of the change from title to name then. ~2026-46260-28 (talk) 22:46, 24 August 2026 (UTC)

• Hatnote. This is in keeping with about two dozen templates that have been designed to represent this sort of information in a standardized fashion. {{Icelandic name}}, {{Spanish married name}}, and {{Bhutanese name}} are a few examples. These German names should be handled similarly. The current footnote produces a paragraph-length description that is far beyond the scope of a biography. It includes details that may or may not have any true relevance to the subject of the article, depending on where they lived, and is actively misleading in cases of Austrian subjects whose titles would have been stripped in 1919. Footnotes are fussy and obstructive, especially alongside bolded names in the lead. If the footnote convention prevails, it should be placed at the very end of the bolded name and the text output should be substantially reduced along the lines of the output of the proposed hatnote template. —Myceteae‍🍄‍🟫 (talk) 00:16, 26 August 2026 (UTC)
• Support hatnote. As Myceteae notes, this approach would be consistent with how we clarify the structures of other non-English names; I also find it clear and concise for communicating this information. ModernDayTrilobite (talk • contribs) 13:51, 27 August 2026 (UTC)
• Support footnote. Prefer over the hatnote because (a) appearance shorter seems better; (b) appropriateness of a footnote at the title as the content is footnote material about the title; and (c) because it would avoid causing change chaos. Transitions are always incomplete and fallible, so when there is enough improvement from a change to justify the issues, then let's just not change. Cheers Markbassett (talk) 16:45, 31 August 2026 (UTC)

  How should the footnote be adapted for the case of Austrian persons? In Austria all titles were banned in 1919. Now the footnote erroneously says that the titles became part of the surname in 1919, as was the case in Germany. Joe vom Titan (talk) 21:08, 3 September 2026 (UTC)

  How are we using it on such Austrian persons? The issue only seems to arise if someone is bold-named as having a title they didn't latterly possess. Which could certain arise because WP:COMMON, but it'd seem a little artless if we didn't also give their post-abolition name too. ~2026-48728-16 (talk) 21:29, 8 September 2026 (UTC)

  The footnote at Template:German title only has one parameter for the title, so it always says virtually the same thing. See the template documentation. these are the pages in Category:Austrian people using the footnote template. Let's look at the first ten:

  • Claus von Stauffenberg was not Austrian, so there the footnote (at the end of the name) is correct. However, it does not explain the title "Schenk". The proposed hatnote explains that title too.
  • Helmuth von Moltke the Elder was not Austrian and died in 1891. The footnote (at the end of the lead sentence) needlessly mentions what happened after 1919.
  • Karl von Frisch moved to Germany around 1910, so the footnote (in the middle of the name) is fine.
  • Georg von Trapp moved to the US and dropped the "Ritter von" from his name, yet we show the full birth name in bold. What happened in Germany is irrelevant to his biography, so the footnote (at the end of the name) is wrong.
  • Karl Philipp, Prince of Schwarzenberg died in 1820, so the footnote (in the middle of the name) shouldn't talk about 1919 at all.
  • Alexander von Middendorf died in 1894. The footnote (just after "von" in the middle of the name) is misused as this guy doesn't have a title. He has nothing to do with Austria.
  • Adam Müller was from Berlin and died in Vienna in 1829. His title is placed in parantheses after his name because he only received it in 1827, the footnote is in the middle of it. Again no need to say anything about 1919 but the footnote does so.
  • Erik von Kuehnelt-Leddihn has his title in the bold name with a footnote in the middle of the bold name. He was born in 1909, so lost his title at age 10. He lived most his live in the US, not sure under what name.
  • Karl Mack von Leiberich died in 1828. He was Austrian, so the footnote (at the end of the lead sentence) mentioning the German situation after 1919 is utterly irrelevant.
  • Cajetan von Felder died in 1894. The footnote is given at the end of his name in German given in parantheses. He was Austrian, so mentioning the German situation after 1919 is utterly irrelevant.

  Joe vom Titan (talk) 12:59, 9 September 2026 (UTC)

  For those Austrian persons who lost their titles early in life, IMHO the bold name should be only their latter name and the title can be mentioned in the Early life section. A few pages do that already e.g. Richard von Mises. Joe vom Titan (talk) 13:08, 9 September 2026 (UTC)

  If we stick with the extremely wordy footnote, then it should be removed from articles such as these where it is erroneous and misleading. I've seen some truly bizarre practices in these articles. An earlier version of Claus von Stauffenberg wrote out his full name Claus Philipp Maria Justinian Schenk Graf von Stauffenberg and included the footnote at the end. The advantage of the hatnote is that it is simple and modifiable. Where additional information is relevant to the subject, it should be incorporated into the article body. For example, Claus von Stauffenberg § Early life and education explains his family background and includes links to Stauffenberg. —Myceteae‍🍄‍🟫 (talk) 15:12, 9 September 2026 (UTC)

  His whole family has the same complicated formatting, e.g. Nina Schenk Gräfin von Stauffenberg. I guess someone reasoned that, if Schenk Gräfin isn't part of her name, it ought to be italicized as a foreign phrase. pburka (talk) 15:27, 9 September 2026 (UTC)

  Dear god. I've cleaned up Nina's article, and Berthold Maria Schenk Graf von Stauffenberg, which looked to be the worst offenders at first glance. —Myceteae‍🍄‍🟫 (talk) 15:45, 9 September 2026 (UTC)

  This also raises questions about usage such as In 2007, Graf von Stauffenberg voiced concerns about the film Valkyrie … We would not repeatedly refer to someone as President Jones or Countess Smith in prose. We should treat it as part of the name or not. If it requires any explanation at all—via any combination of templates, prose, and wikilinks—I would think this should occur once or at most twice. E.g., in the lead and in an 'Early life' or similar section where the family of origin is discussed. —Myceteae‍🍄‍🟫 (talk) 17:57, 9 September 2026 (UTC)

  We should refer to people by the common name in English, even if it is less sensible in the original language. Even if that means repeatedly calling someone "President Jones". – Ike Lek (talk) 23:13, 17 September 2026 (UTC)

  That only works, of course, when there are sufficient recent English-language sources to establish an English common name. pburka (talk) 00:01, 18 September 2026 (UTC)

  True – Ike Lek (talk) 00:57, 18 September 2026 (UTC)

  Agreed, and as far as I can tell the Stauffenbergs are just called "Stuaffenberg". Graf von Stuaffenberg is part of an odd pattern of overusing and drawing maximal attention to these names and titles in several of these articles. I've cleaned up a few. —Myceteae‍🍄‍🟫 (talk) 02:47, 18 September 2026 (UTC)

• Prefer hatnotes to long footnotes, but the hatnote does not need to translate the name. Leave that for wikilinks (and of course article content if relevant there). CMD (talk) 10:02, 10 September 2026 (UTC)

  I agree with this. Explanation that it is a title (and the type of title) in a particular context, and not a name, is enough for the hatnote. How/when that title became a name can (and should) be explained in the linked article. – Ike Lek (talk) 23:10, 17 September 2026 (UTC)

Discussion (German nobiliary titles)

• Tangentially related: Talk:Stauffenberg#Requested move 9 September 2026 —Myceteae‍🍄‍🟫 (talk) 00:09, 10 September 2026 (UTC)

Notes

Can the "link suggestions" newcomer task be disabled for mathematical articles?

Wikipedia has a "newcomer tasks" feature which is designed to "help newcomers make their first successful edits", to hopefully eventually convert some of them to productive editors. To that end, an automated process (i.e. bot) suggests a change, and the new human editor can with a few clicks submit the proposed edit.

See Wikipedia:Growth Team features#Newcomer tasks, mw:Help:Growth/Tools/Suggested edits.

One of these types of edits is a "link suggestion", which seems to work by finding words or phrases that are article titles and appear exactly within some other article's text, and then proposing to the newcomer editor that the word or phrase should be a wikilink.

This feature has been causing continuous headaches to editors of mathematics related articles (I don't have insight about other topics), because a large proportion of the suggested links are either superfluous, mildly out of context, or completely wrong, and even when the links are okay, they are usually of only very marginal benefit. Checking up on every link and reverting the significant proportion of incorrect ones is wasting the time, attention, and good will of experienced editors, and I expect the many reverts are probably also discouraging for newcomers who were just trying to help.

The superfluous links occur where, for example, an elementary jargon word appears incidentally deep into an article about an advanced niche topic. Linking such terms is not helpful to the audience of such articles, as anyone who can make sense of the subject at all is going to have years or decades of familiarity with the basic terms, and the topic of the term per se is not really relevant in context – this is comparable to linking common words in other kinds of articles (say, "apple" or "nothing" or "person"), cf. MOS:OVERLINK. But often the links are outright incorrect, because a phrase which is a concrete jargon term in one context can also be used as separate words with a different meaning in a different context (for example, the phrase "generalized polygon inequality" was turned into "generalized polygon inequality", but this is wrong, the thing being generalized in this case is the inequality, not the polygons).

The basic problem is that the newcomers making these edits do not understand either the text of the article they are editing or the meaning of the term they are wikilinking, and therefore don't (can't) carefully evaluate whether the link is appropriate. I don't know what the link suggestions feature says to the editors using it, but in practice they seem to trust that the suggestions are proper and correct, rather than doing any human checking. So newcomers are basically being turned into a WP:MEATBOT proxy for a rogue automated process.

Can we entirely turn off the link suggestion feature for mathematics related articles and/or links? (As a basic heuristic, we could blacklist any article belonging to the mathematics wikiproject.) I think it's causing more trouble than whatever benefit it is supposed to bring. –jacobolus (t) 16:36, 29 August 2026 (UTC)

Note: There is an ongoing discussion about newcomer tasks and suggested links in general at Wikipedia:Village pump (miscellaneous)#Newcomer tasks and there was a very recent proposal to get rid of suggested links entirely at Wikipedia:Village pump (proposals)/Archive 231#We need to get rid of the "suggested links" tool. —Myceteae‍🍄‍🟫 (talk) 18:38, 29 August 2026 (UTC)

@Myceteae Thanks for the link. From what it looks like, it seems unlikely that there is going to be consensus in favor of turning off the link suggestions feature in the near future. So could we expect to see it be turned off for math articles? As jacobolus pointed out, in that very specific context that feature is a nightmare. Malparti (talk) 19:00, 29 August 2026 (UTC)

Thanks for the pointer. (Before posting I tried searching for "suggestions", since "link suggestions" is the keyword used in edit summaries, and the discussion about "suggested links" didn't turn up as a result.)

It looks like more than a few Wikipedians are unhappy with the link suggestion feature in more general contexts, and find that it routinely makes garbage suggestions. If someone wants to disable or dramatically curtail the tool I wouldn't complain.

(The statistics about revert rate don't seem very convincing to me, on their own: I'm sure many page watchers assume these links should be okay and don't bother checking them, and plenty of the links are moderately unhelpful but I often leave them because it seems like a borderline case, and I don't want to go out of my way to "bite" newcomers. If there's a low revert rate, that might just indicate that a lot of dubious links are being added to articles and then not properly reverted.)

But disabling it for mathematics articles in particular might be an easier or less controversial change. As I said, the basic issue is that the newcomers being asked to do this don't, in general, understand either the context or the term being wikilinked, and aren't spending a lot of time and care, which makes it difficult for them to judge whether the link is correct, and they aren't familiar enough with Wikipedia conventions to know whether the link is appropriate.

If some more general solution is desired, the tool might, for example, include an explicit question next to the link suggestion: "Do you have a good understanding of what this paragraph says, and do you know what «linked term» means?" With a requirement that they affirmatively answer "yes" before being allowed to make the edit. –jacobolus (t) 19:34, 29 August 2026 (UTC)

In reply to Malparti and jocobolus, I suspect editors and articles in other specialized areas face similar challenges. I'm not sure this is a bigger problem in math article than in, say, chemistry or philosophy. That's not to dismiss the concern. A more general intervention might help. Regarding revert rates, my takeaway from these discussions is that the precise figures shouldn't be taken as gospel but these tasks don't appear to create more problematic links overall than we see otherwise. Though that is contrary to some editors' experience. It's conceivable that the newcomer tasks invite editors to dense, technical articles that they were unlikely to stumble across or feel inclined to edit on their own. I hadn't paid attention to newcomer tasks until recently so I'm drawing from what I've read in all these recent discussions along with my own experience of MOS:OVERLINKing, which is a pervasive problem. —Myceteae‍🍄‍🟫 (talk) 20:05, 29 August 2026 (UTC)

One of the biggest problems is the asymmetry in the burden placed on experienced editors. You have someone who doesn't know or care anything about a topic per se spending a few seconds to confirm a machine-generated edit. Then if the edit is obviously bad, you have 1–2 experts checking it for a minute each and making the revert. If the edit was borderline or good, and the decision is made to not make the revert, you might force an additional 5–10 experts to spend a minute each evaluating the change (there's no way to mark an edit as "this was checked and seems okay"). Plus some overhead, and the distraction of cluttering up their watchlist with stuff that is mostly irrelevant to the projects they care about.

In the best case, the benefit we get is one additional wikilink that with high probability will never be clicked, or at most will be clicked a few times by readers over the lifetime of the page. These newcomer editors aren't being obviously converted to competent writers and experts who can write or make major changes to math articles, so for math articles in particular the side benefit from recruitment is slim to nothing.

Overall, it's not a good way of respecting experts' time. Plenty of the article watchers making reverts here are university professors with PhDs who have lots of other responsibilities, and are volunteering a small bit of their time to working on Wikipedia as a public service. But instead of optimizing their time use getting them to write new articles, we're squandering it with busywork. –jacobolus (t) 20:47, 29 August 2026 (UTC)

I get these concerns but the link task is the least disruptive of all of them as it doesn't deteriorate the actual prose, whereas everything either deteriorates the prose or inserts junk citations or both. Gnomingstuff (talk) 21:02, 29 August 2026 (UTC)

In mathematics articles we don't seem to see many other "newcomer task" edits. Perhaps because the content is technical enough that newcomers with no relevant expertise are discouraged from making changes? So the link suggestions are the ones that are most obviously obnoxious. –jacobolus (t) 21:06, 29 August 2026 (UTC)

That might be part of it. IIRC people choose the (extremely broad) subject matter area and then get arbitrary articles shoved at them to spam out edits to Gnomingstuff (talk) 06:23, 30 August 2026 (UTC)

The most surprising thing here is the claim that the English Wikipedia actually has up to 10 "experts" checking each edit to a math article. I doubt that's true. WhatamIdoing (talk) 19:29, 1 September 2026 (UTC)

You are disputing that the people who watch math pages are experts, or that they try to check up on miscellaneous changes, or just that there are more than a couple of people who care about the correctness of Wikipedia math articles? –jacobolus (t) 19:56, 1 September 2026 (UTC)

I'm surprised that so many are available in practice. In fact, I doubt that every edit to math articles gets checked by 10 editors at all, much less by 10 editors who understand the subject area. (One can care very deeply, and still have real-world factors that prevent spending all day on wiki.) WhatamIdoing (talk) 21:41, 1 September 2026 (UTC)

The suggested links tool is just a step towards having an AI take over editing of the encyclopedia entirely. Today it is using newbies as proxies, tomorrow it will be getting rid of the middleman altogether. As it stands, any editor, newbie or not, who follows a suggestion to make a clearly wrong link raises WP:COMPETENCE questions. BD2412 T 20:39, 29 August 2026 (UTC)

I sincerely hope that the WMF is not dragging us down that route, but their repeated secretive introduction of AI against our clearly communicated consensus makes it really difficult to continue assuming good faith. Certes (talk) 20:57, 29 August 2026 (UTC)

Sorry, but "sincerely hope" usually implies that deep down you know the tsunami is coming but hope it will not. I fully agree with you that it would be a disastrous decision by WMF, but accept that they will do what they like and give us some type of word salad to justify it. Sorry, but that is how it is. Yesterday, all my dreams... (talk) 20:23, 1 September 2026 (UTC)

Has Musk secretly taken over WMF? Who knows? Yesterday, all my dreams... (talk) 20:41, 1 September 2026 (UTC)

The issue is not limited to mathematics articles. A recent edit linking machine to machine when the process was manually carrying a tray of cards from one machine to another. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:31, 31 August 2026 (UTC)

If there's a template on most of these articles, Template:No newcomer task could be added to the template, and thus propagate to all the articles transcluding it. WhatamIdoing (talk) 18:43, 1 September 2026 (UTC)

I think this reeks of elitism. Only those who already know the magic incantations are allowed to participate? People are slowly turning this encyclopedia back into Nupedia. Let people make mistakes. Explain it to them and move on. It doesn't matter if they used this tool or not, it's just people wanting to participate and trying to figure out how to do that. That's the wiki way and the only way a person becomes an editor to replace you (generic you), who are approaching the age of no longer being part of this encyclopedia either. This has nothing to do with competence, and everything with people seeking a perfection that doesn't exist. —TheDJ (talk • contribs) 18:58, 1 September 2026 (UTC)

I have no idea what you are trying to say. Magic incantations?

The problem is not "people". The problem is having many wikilinks generated by a "machine learning" tool which is incorrect a large proportion of the time, and even when correct only marginally helpful, with the result that it wastes a whole bunch of human time and attention for not much benefit.

If human readers come across an article "organically", are reading along and think that a wikilink is missing, so decide for themselves to add it, nobody has a problem with that; even when newcomers often make mistakes or don't understand conventions, generally experienced editors are happy to help. Absolutely nobody is saying that new users should be disallowed from participation. –jacobolus (t) 19:24, 1 September 2026 (UTC)

Second onboarding screen in Mediawiki "add a link" workflow using the term "machine".

You asked above about what the newcomers are told. Here's a screenshot for one of the instruction panels.

The problem is, in fact, "people". Specifically, the problem is that one group of people is new (and nobody's good at this stuff when they first start), and another group of people (experienced editors) have personal preferences that vary significantly, resulting in differing opinions about whether to add or remove a given link.

For example, how many long-time editors know that WP:OVERLINKING got redefined a few years ago as more than one link per section, rather than the old standard of more than one (or rarely two) links per entire body-of-the-article? This results in some editors claiming OVERLINKING violations for links that comply with the guideline.

We also see experienced editors who think that everybody knows ____ (or at least the most typical/intentional readers of the article), so it's just too well-known to justify a link to that basic concept/tiny country/broad article. For example, does Exponentiation need the link to Addition that you reverted? I don't think so. Is it revert-worthy? I don't know. But perhaps relevantly to this particular discussion, I know that the edit that added that link wasn't a Newcomer task. It was added by a 10-year-old account with >3,000 edits. WhatamIdoing (talk) 19:54, 1 September 2026 (UTC)

Yes, we have been having a discussion about that page; after making that revert (to a large edit that made a big bundle of unrelated changes) I immediately started a talk page discussion. I had been in the process of restoring about half of the changes I reverted (as I said in discussion, a few examples gave me an unfairly bad impression of the whole thing, and a wholesale revert was perhaps unjustified), but my re-revert edit got into a significant edit conflict and I had to go do a real-world errand. Some of the reverted changes were restored by the other editor, while others were not. I need to go back carefully through and figure out which of their changes are helpful, which ones that were reverted should be restored, which that were restored should be discussed, etc. This is an example of the Wikipedia process working as intended: a human editor makes a good-faith change, another human editor makes a good-faith revert, and then we have a discussion to establish consensus. The whole process takes a lot of effort: checking, considering, discussing, persuading, writing and rewriting. But the end result is hopefully better than what we started with.

That's entirely unrelated to the topic of machine-generated changes. –jacobolus (t) 20:02, 1 September 2026 (UTC)

(I wish that partial reversions were easier to do. Maybe WMDE's work on the edit conflict tool could be adapted to make line-by-line choices about what to revert possible?) WhatamIdoing (talk) 21:13, 1 September 2026 (UTC)

To the point of the newcomer task tool though:

This wording clearly isn't sufficient in practice to get the newcomers to actually follow the direction to "use your judgment to decide whether they are right or wrong". People routinely implicitly trust that what they are directed to do is correct and justified (even to the point of taking clearly unethical actions, cf. Milgram experiment). I think you put far too much faith in a couple words of instructions, easily skimmed past. Has anyone done explicit user testing of this? (No, just gathering summary statistics doesn't cut it.)

Throughout this discussion, you keep flippantly excusing a tool that is, in practice, causing significant annoyance to human Wikipedians. It feels frankly quite rude. –jacobolus (t) 20:10, 1 September 2026 (UTC)

@WhatamIdoing Instead of imposing this attention cost on innocent Wikipedians (whether or not you think they are "expert" enough that their time should be valued), how about you can personally volunteer to double-check every machine-generated "newcomer edit"; the edit can be temporarily blocked from application until you have verified that it is correct, then we can let it through. Or if you don't want to spend your own time, maybe you can convince the WMF to pay someone a fair wage to do the checking, so we don't waste volunteers' time until at least one vaguely competent person has checked it. –jacobolus (t) 20:18, 1 September 2026 (UTC)

Why do you think that having a newcomer make an edit should be understood as "imposing this attention cost on innocent Wikipedians"? Did no "innocent Wikipedians" look over your own early edits? Let's see: your first edit was to remove someone else's photo and swap in your own. (It's a nice photo; thanks.) It looks like your second edit added about 10 sentences of unsourced and sometimes opinionated content about a game. Your third edit in the mainspace added a link to Jeans – the second link to that article in that same section. What are you having to do for these newcomers that wasn't done for you? WhatamIdoing (talk) 21:08, 1 September 2026 (UTC)

your own early edits?

As has already been repeated ad nauseam, nobody is mad about newcomers' "own early edits". The concern is with a bad "machine learning" algorithm which is using newcomers as a meatbot proxy for unhelpful changes.

If you want to go spend your time reviewing my Wikipedia edits from when I was a college student >20 years ago, you are welcome to do so, but it is entirely irrelevant and off topic to this discussion. Maybe you can put your critiques on my talk page instead, or you can go manually revert any parts that still persist to today that you think are bad. –jacobolus (t) 22:11, 1 September 2026 (UTC)

This system is how many promising newcomers make their first edits. Most of the edits aren't "unhelpful".

I looked through your most recent 100 mainspace edits. I found several examples of you reverting edits by newcomers (and other editors, e.g., ). But in the last ~two weeks/100 article edits, it appears that you edited just three (3) articles as a result of the newcomer tasks, two of which were reverts (links to quadratic equation and fourth power – IMO reasonable choices that you decided weren't important enough to link to), and the third of which was you refining a correct link to a redirect to the same article.

The math's not working for me here.

• If you reverted two edits, and most (>50%) of these edits are bad, then you couldn't have seen more than three of these edits recently.
• If exactly half of the edits are bad, then you have seen just four of them.
• And if most of them are good, then why are you complaining here?

I wonder whether you are incorrectly blaming the Newcomer tasks software for unrelated edits by newcomers. If you think you're not, then I wonder if you could tell us what percentage of correct suggestions would be necessary for you to feel like it wasn't terrible?

Or is the problem really just that newbies are making edits that light up your watchlist, and you wish they'd leave "your" articles alone? WhatamIdoing (talk) 00:59, 2 September 2026 (UTC)

Your comments continue to feel really rude and personalized. Since you don't like the message, you are doing everything you can to shit on the messenger.

I came here to make this post because several other people were complaining at the Math Wikiproject talk page. It's causing significant distraction and annoyance. People are frustrated. Judging by the comments of others here, it's also not just editors of math-related articles who think there's a problem. You don't need to deny that experience or feign incredulity. If you don't think that the edits are a problem, I ask you again: why don't you put your time where your mouth is and personally volunteer to double-check them all before they get applied? –jacobolus (t) 01:07, 2 September 2026 (UTC)

I can't "double-check them all before they get applied" because there's no way for me to do that. Edits made by newcomers cannot be checked until after the edit has been made. WhatamIdoing (talk) 01:15, 2 September 2026 (UTC)

There's no way to do it as the feature is currently implemented, but it's certainly physically possible to implement it differently.

There's currently a Wikipedia:Pending changes feature that blocks certain edits from being displayed until after they have been reviewed. Something similar to that could be put into place for these "suggested links" edits, and a team of self-selected patrollers could be responsible for reviewing them. We could tell everyone else to just ignore those edits and hide them from their watchlists until they had been reviewed. –jacobolus (t) 01:25, 2 September 2026 (UTC)

The problem is, in fact, "people".

This. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 00:18, 9 September 2026 (UTC)

Jacobolus, you are 100% correct, alas about 5% likely to succeed on this given the nature of consensus on semi-major decisions. I would have also suggested physics, not just math. Any way good idea, but C'est la vie on Wiki. Yesterday, all my dreams... (talk) 20:15, 1 September 2026 (UTC)

Ping @KStoller-WMF – Is it possible to disable this tool for math articles? Or is there some other way that the tool can try not to propose articles/wikilinks to new editors who understand neither the text of the article nor the meaning of the linked term? Or that the tool can be rewritten to make a much lower proportion of unhelpful links? –jacobolus (t) 20:24, 1 September 2026 (UTC)

To avoid links to complex articles you would need a "measure of complexity" for the article. It would be very useful to have that, but also quite difficult to implement automatically. Let us hope they will not try to use LLM for that. Yesterday, all my dreams... (talk) 20:29, 1 September 2026 (UTC)

Another idea: perhaps a scraper bot could check up on every "link suggestion" ever made a few times (maybe 1 day, 1 week, 1 month, and 6 months after the initial edit), and for every one where the link no longer remains in the article, could: (a) blacklist that page title from ever be proposed again as a "link suggestion", and (b) could compile a list of other articles where the same link was added but wasn't reverted, so that human editors could go double-check on edits which have a pretty good chance of being wrong. –jacobolus (t) 20:40, 1 September 2026 (UTC)

Good, you have just described the first steps in a machine learning approach to doing it. Yesterday, all my dreams... (talk) 20:44, 1 September 2026 (UTC)

Ostensibly this whole feature is "machine learning", but there doesn't seem to be much learning involved in the current version, which is more like "machine keeps making the same mistake over and over despite being corrected". –jacobolus (t) 20:46, 1 September 2026 (UTC)

@Jacobolus to answer your specific question, yes I believe Category:Mathematics could be added to the exclusion list for Add a Link task. That's actually already something that is Community Configurable, and an English Wikipedia admin can update the "Articles containing categories defined here will not be shown to users as tasks for this task type" field via Special:CommunityConfiguration/GrowthSuggestedEdits if it seems like there is community agreement that link suggestions on Mathematics articles are particularly problematic.

Yesterday I shared a summary of some of the recent & upcoming improvements for Add a Link: Wikipedia talk:Growth Team features#Newcomer Tasks / Add a Link improvements. The TL;DR is that we released a model improvement for Add a Link yesterday, other improvements are planned, and a significant change will be released ASAP: T429417 Add Link should not make a recommendation if one has already been accepted or declined, which will limit each article to a single Add a Link suggestion ever.

I'll be direct about where I stand: I'm taking the concerns in this thread seriously, and attempting to prioritize some quick improvements, and I also still believe newcomers need easy entry points into editing. I hope we can find a balance that reduces the cleanup burden for experienced editors while keeping the door open for the next generation of editors. - KStoller-WMF (talk) 23:49, 1 September 2026 (UTC)

Why should the "easy entry point" for a human be enacting a decision made by a machine process?

In my opinion, if we have a bot that we think is correct with a very high likelihood (say, well above 99%), and a consensus among Wikipedians supports its operation, then we should just have the bot make the edit. If we have a bot that we think has a high chance of being incorrect (certainly anything more than 5% errors), we should scrap the bot as being inadequate. Having a bot that has a high chance of being incorrect, and then choosing completely inexperienced passers-by (who with high likelihood haven't even read the article to understand the context) to review the changes is just a recipe for bad edits and frustration by folks who end up having to clean up after. It's also a completely synthetic activity that bears very little resemblance to what we hope those people will do later, if they decide to stay around. Is there any evidence that performing this link suggestion task is effective for recruitment of folks who otherwise would have left, but instead become productive members of the community?

If you want to show people that they can make changes to the wiki, it seems to me like the first step is to help those human editors personally identify a problem, and then show them how to go about addressing it.

Rather than foisting new editors off on a completely impersonal machine-generated idea, we should be trying to help those newcomers more direct feedback/interaction with more experienced human editors. –jacobolus (t) 00:29, 2 September 2026 (UTC)

We should care about an easy entry point for new people because I am going to die. The "old hands" will not live forever. WP:OBIT gets longer every year, and you have already outlived some of your fellow editors. We need to recruit the next generation of editors. WhatamIdoing (talk) 01:05, 2 September 2026 (UTC)

Can you point to a single productive author of math-related Wikipedia articles who started with a "link suggestion"? Did you go ask them directly if the link suggestion made them more likely to stick around?

The supposed benefit here seems 100% hypothetical, I don't see any evidence that it is a real non-trivial thing. So basically you're celebrating a concrete and clearly expressed actual problem out of a vague, frankly unjustified, hope about the future. –jacobolus (t) 01:14, 2 September 2026 (UTC)

How many productive editors of math-related Wikipedia articles can you point to, who have been editing for less than, say, 18 months? Because 20 months ago, nobody here had access to this feature, and 18 months ago, only a small fraction of new editors could see it. WhatamIdoing (talk) 01:20, 2 September 2026 (UTC)

I thought it'd be useful to put some numbers on this. quarry:query/108882 says that over the course of six months earlier this year, there were:

• a mean average of 11 add-a-link edits to articles tagged by Wikipedia:WikiProject Mathematics per day,
• of which just 4% got reverted (i.e., about three times a week).

This does not seem like an overwhelming flood of edits to me, nor an unusually high proportion of revert-worthy edits from newbies. Even if someone had all 33,000+ WPMATH articles on their watchlist, this would still be a small proportion of math-related edits and an even smaller number that needed to be reverted.

If there is a big problem with bad links being added to math-related articles, I suspect that this is not being done by newcomers using this tool. WhatamIdoing (talk) 04:26, 2 September 2026 (UTC)

Okay, so there were about 2000 edits. Can you go check them? It sounds like there are potentially another hundreds of errors there. –jacobolus (t) 05:21, 2 September 2026 (UTC)

What makes you believe that there could be "hundreds" of "errors" in those articles, when only a total of 82 were reverted during the entire first six months of the year?

I could get a list of diffs, but we appear to have different views on what links are desirable. I wouldn't have reverted the link to quadratic equation that you did, and I might not have reverted the link to fourth power. Consequently, I doubt that you would trust my review of them. WhatamIdoing (talk) 00:04, 3 September 2026 (UTC)

Arguably Fourth power shouldn't be an article at all, since it's more or less just a combination of the number 4 and the concept of an exponent, and we don't really have much of anything special to say about it as a particular case. But leaving that aside, the link to "fourth power" is clearly inappropriate in the context of the sentence This is because of the Rayleigh-Jeans law, which states that at frequencies much lower than the peak frequency of a black-body radiator, spectral power is inversely proportional to the fourth power of the wavelength. The wikilink says nothing remotely relevant to this text; if people are curious about why the quantity ⁠$\lambda ^{-4}$⁠ appears in this equation, they are much better off reading Rayleigh-Jeans law. The link to quadratic equation far down the article Pythagorean triple is also completely inappropriate; the article is full of polynomial equations, many of them quadratic, and the use of the term here is completely incidental; nothing at the wikilink is directly relevant in context – in particular, the article at quadratic equation is entirely focused on equations of one variable, but the equation being described in the relevant sentence of Pythagorean triple has 4 variables.

What makes you believe that there could be "hundreds" of "errors" in those articles, when only a total of 82 were reverted during the entire first six months of the year?

What makes you think there wouldn't be hundreds of errors? Every incentive is to ignore these edits, or, even for folks checking them, to leave them alone unless they are completely egregious. I imagine most of the 2000 edits were never carefully considered. –jacobolus (t) 00:29, 3 September 2026 (UTC)

I notice that today you tell me that "Every incentive is to ignore these edits" and "most of the 2000 edits were never carefully considered".

On the other hand, less than 48 hours ago, you were telling me that "it wastes a whole bunch of human time and attention" and that we're "imposing this attention cost on innocent Wikipedians", and two days before that, you were telling me that if editors don't revert all these edits, then they "force an additional 5–10 experts to spend a minute each evaluating the change".

It is not possible for both of these stories to be true. Either most of the links aren't being "carefully considered" or we don't have a bunch of expert editors collectively spending 10 minutes looking at that each link. It is not credible to claim that "expert" editors can spend that much time looking at a single link without meeting an ordinary standard of "carefully considering" the link. After all, most editors can evaluate the correctness of most links in just a few seconds.

Would you like to pick one of these mutually exclusive stories? Either it's taking a inappropriate amount of time and attention for multiple editors to review these links – in which case, it logically follows that the existing revert rate is correct – or nobody's looking at them – in which case, it logically follows that the math editors aren't being forced to waste their time on these edits. I don't really care which one it is, but I'd like you to commit to one of them. If you can do that, then I'd be happy to do what I can to find out whether there's any reason to believe your chosen story is true. WhatamIdoing (talk) 01:25, 3 September 2026 (UTC)

Both stories are simultaneously true:

(1) It takes an inappropriate amount of time and attention to actually review these edits properly. To the extent they get reviewed it's an annoying burden.

(1b) Because there is no way to mark a particular edit as reviewed (other than by reverting it), while edits that get immediately reverted are probably only noticed or checked by 1 or 2 people, edits to highly watched pages that someone decides not to revert might be checked, or at least skimmed, by several other people; I speculated 5–10 before, but there's no actual way to count since the data is not collected anywhere.

(2) Because there have been many of these edits, including many to very obscure pages with few active page watchers, a large proportion of the edits are not getting carefully checked. In some cases someone might give it a brief glance, but without carefully reading the context and carefully checking the content of the wikilink. Many edits which are therefore adding incorrect or inappropriate links are going to slip past. You user:WhatamIdoing provided a very good example of what can go wrong: even when you tried to do a check of some edits that were explicitly reverted, because you didn't actually look/think carefully about them, your immediate impression was (wrongly!) that the edits were fine. This surely happens often, leading to a high proportion of mistakes. –jacobolus (t) 05:49, 3 September 2026 (UTC)

In fact, WhatamIdoing, we can notice something stronger from your experience: even a 20-year Wikipedia veteran can't accurately evaluate the relevance of the machine-generated wikilinks when they are looking quickly. And yet, this entire feature is premised on getting complete newcomers to do so! How can a brand new user do a good job at this task if even someone like you can't? –jacobolus (t) 07:59, 3 September 2026 (UTC)

Your argument is built on the unproven claim that I'm wrong about those links. WhatamIdoing (talk) 03:13, 5 September 2026 (UTC)

There are 4 relevant claims involved: (1) The specific wikilinks are not particularly relevant to the context where they were put; this is a fairly straightforward and uncontroversial claim, which we can verify by examining the content of the wikilinked articles and the context where the links were added. (2) Readers who clicked these wikilinks would probably not glean that much useful context about the article they were reading; this is subjective and we don't really have a way to easily test it. (3) Including these wikilinks is on balance unhelpful for the articles where they were put; this is a matter of personal opinion and taste. (For example, someone might make the argument that even if the current article Quadratic equation doesn't at all address 4-variable equations like the one being described by the phrase where it was linked, an article with the title "quadratic equation" could plausibly be rewritten to prominently discuss the more general topic of quadratic equations in any number of variables, and in a future world after that rewrite, the link might eventually become relevant. I don't personally think this is a reasonable basis for adding wikilinks to substantially off-topic articles, but there's no way I can force the other person to agree with me about that.) (4) If polled after a discussion and careful examination, these wikilinks would be rejected by community consensus of the editors who commonly work on math articles; this one is an "unproven claim" but not hard to test: we can directly ask at the math wikiproject if you want. –jacobolus (t) 03:52, 5 September 2026 (UTC)

WPMATH is not "the community", and it is not only "the editors who commonly work on math articles" who are allowed to form a consensus.

I dispute your (1), as a wikilink that is "not particularly relevant" to you could still be helpful to someone else, e.g., if they are not a native speaker of English.

I would add: (5) the specific wikilinks pointed to the correct article; this is undisputed and very important.

So that's 1, disputed; 2, unknown; 3, personal preference; 4, an appeal to like-minded authorities, and 5, undisputed and in favor of the link. For me, that adds up to the link being acceptable. WhatamIdoing (talk) 04:04, 5 September 2026 (UTC)

What do you mean "pointed to the correct article"? I don't know what you mean by "acceptable". Like, would I recommend banning someone for adding such a wikilink? No. Would I revert a change that added it, and start a discussion to establish consensus if the other person disagreed? Yes. As for "like-minded authorities": who do you expect should decide what wikilinks should be included in math-related Wikipedia articles other than the authors and maintainers of math-related Wikipedia articles? Overall, this is one of the more absurd lines of argument I have ever seen put forth on Wikipedia, which is really saying something. –jacobolus (t) 04:08, 5 September 2026 (UTC)

One of the problems that we've seen with newcomers adding links is when they add links to "John Smith", but we need them to add a link to "John Smith (athlete)". Fourth power was the correct article to link, but sometimes people end up linking to the wrong article (e.g., to any article listed in Fourth power (disambiguation) or the business named Fourth Power).

WPMATH ≠ the authors and maintainers of math-related Wikipedia articles, and our WP:LOCALCON policy exists because of a WikiProject deciding that "their" articles should be exempt from ordinary MOS rules. WPMATH does not have the right to determine whether other editors are allowed to follow MOS:LINK. WhatamIdoing (talk) 20:26, 5 September 2026 (UTC)

I don't understand what you are trying to say.

Nobody is claiming that math articles are "exempt from ordinary rules". MOS:OVERLINK is part of the manual of style (a "guideline", i.e. a "set of best practices supported by consensus"). Its guidance covers such cases: "words and terms understood by most readers in context are usually not linked". What counts as "understood by most readers" depends on the context of the article; for technical sub-sections deep into specialized mathematical articles the "most readers" in question are a much smaller group than generic readers of Wikipedia articles about topics of broad interest, so how to apply this style guideline is a matter of judgment by Wikipedians working on topical articles.

MOS:UNDERLINK says that what should be linked is: (1) "Relevant connections to the subject of another article that help readers understand the article more fully", (2) "Articles with relevant information", and (3) "Articles explaining words of technical terms, jargon or slang expressions or phrases", and (4) "Proper names". None of these is really applicable to the links I reverted. The reason I reverted these links is because the linked articles do not contain information substantially relevant to the context.

I propose WT:WPM as a place to gauge editor consensus because it's a place where you are likely to find Wikipedians who care at all about these topics, would be willing to take a look, can make easy sense of the wikilinked article and the context where the link was added, and can apply competent judgment about whether the link is relevant and appropriate.

If you can find a different way to evaluate consensus among "authors and maintainers of math-related Wikipedia articles" (which you seem to think is a distinct group from WP:WPM), then suggest away. –jacobolus (t) 21:20, 5 September 2026 (UTC)

No, none of these is applicable to the links you reverted in your personal opinion. In the personal opinion of other editors, those links are appropriate.

WT:WPM is a good place to gauge editor consensus, if you define "editor consensus" as "the consensus of only a small, self-selected, and non-representative group of editors". If one wants only people editing math-related articles, then there's a list at Wikipedia:WikiProject Directory/Description/WikiProject Mathematics#Active Subject-Area Editors. But the real goal is "editors", which includes those who don't know a lot about math or normally edit in that area. WhatamIdoing (talk) 16:24, 6 September 2026 (UTC)

"In the personal opinion of other editors, those links are appropriate." – Which other editors are we talking about? Whichever newcomer supported adding these machine-generated links probably gave the matter only cursory consideration, and the same was presumably true for you when you offered them as examples.

"includes those who don't know a lot about math" – Why do you want editors who "don't know about math" to be judging the relevance of wikilinks buried deep in technical sections of math articles? They aren't the target audience as readers, don't care about the topic, and often can't make sense of either the wikilinked article or the context where the link is added. –jacobolus (t) 17:00, 6 September 2026 (UTC)

Why should the "easy entry point" for a human be enacting a decision made by a machine process?

Well, the Growth Team features/Newcomer tasks could help clear the backlog of articles that are tagged with CN ("add a citation") or "needs copy edit"/"promotional?" ("revise tone"), even if these suggestions are machine generated rather than, say Category:All articles lacking sources or Category:All Wikipedia articles needing copy edit. Sure, they could disable just the "Add a link" features, but certainly not all of the newcomers tasks. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 00:35, 9 September 2026 (UTC)

@User:KStoller-WMF probably should be in the Growth Team talk page, but I have a suggestion: add all the categories in the Help Out section of the Wikipedia:Community portal into the newcomer task system? Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 00:38, 9 September 2026 (UTC)

Well, the Growth Team features/Newcomer tasks could help clear the backlog of articles that are tagged with CN ("add a citation") or "needs copy edit"/"promotional?" ("revise tone")

This is already happening. There's a parallel discussion going on at Wikipedia:Village pump (miscellaneous)#Newcomer tasks about how this "help" is equally unhelpful.

My sympathies with all the other editors who have now collectively tried dozens to hundreds of ways of explaining the problem in the feeble hope that maybe, just maybe, one of them will actually get through to people. Gnomingstuff (talk) 06:04, 9 September 2026 (UTC)

Rather than a "measure of complexity" for the article, I think it would have to be a measure of complexity for the subject. Freudenthal algebra is a simple "article", but a complex "subject".

I believe that the current system looks first for a density of links that is below average. Most articles have 10–50 wikilinks, or about one wikilink for every 20 words. If you don't want the system to suggest adding links to an article (and for whatever reason, you can't put {{No newcomer tasks}} on the article, which blocks it entirely), maybe try to make that article look like it doesn't deserve an {{underlinked}} tag. That should have the effect of making the article invisible to the newcomer task system. WhatamIdoing (talk) 21:31, 1 September 2026 (UTC)

I didn't realize "No newcomer task" was a thing. Should we just blanket add that to all technical articles? –jacobolus (t) 22:13, 1 September 2026 (UTC)

I mentioned it above. I don't think that blanket application to hundreds of thousands of articles is a good idea. WhatamIdoing (talk) 00:27, 2 September 2026 (UTC)

A simple bot could do it, but would require an Act of Congress, so to speak. Yesterday, all my dreams... (talk) 04:54, 2 September 2026 (UTC)

Yes. Stepwise Continuous Dysfunction (talk) 05:04, 5 September 2026 (UTC)

I sometimes feel like we are being trolled with a big practical joke - "let's build a machine to trick newcomers into making nonsensical links throughout Wikipedia, and then get earnest Wikipedians to defend this machine". BD2412 T 17:50, 9 September 2026 (UTC)

The feature seems to work well with Military History articles. I have reviewed a score of them and only once found an inappropriate link. I realise that readers frequently don't follow the links - maybe many of them don't even realise that they exist - but getting newcomers to realise that they can edit the pages is an important step. Hawkeye7 (discuss) 05:18, 4 September 2026 (UTC)

I'm late to the party, but I feel that worrying about the quality of links in maths articles is a bit like worrying about whether your toenail varnish is the right shade when you've just broken your leg. This is an area of Wikipedia rife with hopelessly over-complex articles that make no attempt to put the subject in encyclopedic context, and instead look like they were produced by a grad student wanting to demonstrate that they know more about the itty-bitty details of their subject than anyone else (and that they can format equations more prettily). The maths articles are in dire need of people who can differentiate between a text book, a primary publication, and an encyclopedia, and who can write for an intelligent non-expert. Elemimele (talk) 16:47, 9 September 2026 (UTC)

Yes, which is why it's a big problem to have a bunch of the people who could be (that is, routinely do) rewrite math articles to be significantly better instead spend time checking up on bogus links generated by a robot. –jacobolus (t) 17:00, 9 September 2026 (UTC)

Assuming that "checking up on" 11 links per week has a significant effect on the workload of all our maths editors. I just checked the last 11 edits for suggested tasks. It took me an average of 30 seconds each, including reverting one, changing another, and creating a redirect. The rest were either good (most of them) or unimportant (do we really need a link to WWII? Eh, it doesn't matter). Even if we assume that math articles are three times as complicated, we'd still only be talking about two minutes per day. But if it's the same, you're whingeing about people "wasting" an average of 45 seconds per day, by making you look at links that are almost always (>90%) kept in place. How many math articles do you think you can rewrite in 45 seconds per day? WhatamIdoing (talk) 18:44, 9 September 2026 (UTC)

I dispute that it only takes 45 seconds to check each of these links. You need to at least read the paragraph where the link appears, click the wikilink, and skim the wikilinked article to check what it contains. Sometimes it's possible to immediately conclude the link is good or bad, but other times it might quite a bit more attention and effort than that. Especially if someone e.g. leaves a talk page message for the new editor explaining their decision.

This is clearly the type of activity that you personally enjoy, but many other people don't really. This is why I say you should organize a team of volunteers (or paid staff) to do this cleanup, with these edits gated behind the check, instead of foisting it on innocent Wikipedians.

Every bit of random context switching takes significant attention, beyond the precise time it takes to do. It jolts people out from whatever task they were otherwise planning to do. (As a concrete example, four things I didn't really want to be working on – (a) replying to your comments here, and (b) dealing with the WMF's imminent math rendering change which is expected to break math formulas on most English Wikipedia articles, (c) trying to clean-up after a well-intentioned newcomer who changed a bunch of math history articles to reflect their personal views as cited to some self-published presentation slides they made, and (d) checking changes to articles I am watching – have taken up most of my Wikipedia editing time and attention in the past few days, preventing me from working on the research and writing of the couple of articles I actually wanted to improve.)

In some cases (e.g. fighting vandalism, helping clean up after well-intentioned newcomers who don't yet know Wikipedia conventions, hunting for sources for obvious and uncontroversial claims after someone asks for one on the talk page) an attention penalty is probably unavoidable, though as a community we could probably do better if we put more effort into considering systems/processes. In some cases (e.g. cleaning up messes left by Citation Bot) the attention penalty should be avoidable, so editors are routinely frustrated, but do their best. In this particular case, the cost doesn't seem to be justified by any apparent advantage at all, at least to math articles. We're turning experts into janitors for a robot that makes only trivial improvements with lots of mistakes mixed in.

If you get someone annoyed and they feel like their time and efforts are being abused, and they then leave, they might not write any math articles at all. At best, if you distract them with enough stuff unrelated to their previous train of thought, their progress is going to be significantly slowed down. –jacobolus (t) 21:34, 9 September 2026 (UTC)

It's the nature of an average that some are bigger or smaller, but I reviewed 11 actual edits, with a stopwatch running, and it took me an average of 30 seconds each. I did not rush. I did have to open some of the links to make sure they were correct; I did have to take time to correct or revert a couple. But it did not take much time.

If you don't like doing this work, then you aren't required to WP:VOLUNTEER to do it. You can ignore it. WhatamIdoing (talk) 19:34, 10 September 2026 (UTC)

As far as I can tell none of the people doing this work for math articles do it because they like it. Nobody wants to be spending their time cleaning up after incorrect or unhelpful machine-generated wikilinks. (Or cleaning up after Citation Bot, or cleaning up after vandals; etc.)

But people want some set of articles to remain correct, clear, accessible, etc., so they feel some obligation to check on changes to those articles and revert mistaken ones.

When people complain, here you come to tell them: "you care about articles now, but have you considered just not caring?" But this is mostly evidence that you personally have not even tried to engage with people's criticism or personal motivations. Instead you are consistently dismissive and frankly disrespectful. –jacobolus (t) 19:49, 10 September 2026 (UTC)

Where's your evidence that it takes significantly longer than about 45 seconds a day, or five minutes a week, to check these links?

I'm not trying to dismiss the concern. I'm trying to inject a little bit of reality. There are more than 33,000 math articles. This newcomer task has produced just one or two edits to them each day this year. One or two edits per day is not a lot, right? Even if you have to do things like "read the paragraph", it's still not a lot of time as measured on a clock, right?

How many hours have you spent trying to convince me that these edits are a burden on your time? How many months' or years' worth of edits could you have reviewed during those hours? WhatamIdoing (talk) 01:33, 11 September 2026 (UTC)

It seems appropriate to put some numbers on this:

Using an old list, I've found that there have been 110 newcomer-task edits to WPMATH's articles during the last week. That's 15 edits per day, including all newcomer tasks, not just the add-a-link task. Exactly 10% have been reverted.

Using that same list, I've found that there were 751 edits to WPMATH articles by newer (<500 edits/non-extended confirmed) registered editors during the last week that weren't using a newcomer task. That's 110 non-newcomer-task edits per day. Almost 12% of them have been reverted.

Newcomers at math articles

Type	Raw count
Organic edits	751
Newcomer task	110

Reverts at math articles

Type	Raw count
Organic reverts	87(12%)
Task reverts	11(10%)

This tells me:

• The newcomer task tool is only a source of a small number of edits. If you randomly select an edit, it's probably not an edit from a newcomer task.
• Newcomer task edits have a lower chance of being reverted, compared to new editors making edits on their own.

Looking more specifically at individual edits:

Five of the reversions were links to If and only if (a problem that cropped up recently and has been discussed at WT:WPMATH). Would you agree with me that those probably took just a couple of seconds to revert, and therefore did not represent a significant time sink for maths-focused editors?

So that leaves us with just a few newcomer edits that got reverted during the last week that might have taken more than a few seconds to make a decision about. Some of those weren't add-a-link edits anyway, and so can't be blamed on the add-a-link task. Here are the reversions that might have taken more than a few seconds: (wrong link), (preference for fewer links, citing the brand-new and WP:PROPOSAL-free recent change to MOS:MATH). And one editor moved a link to the top of the article, which was counted as a reversion but really wasn't.

There are many more non-newcomer edits that got reverted (87 vs 11) – some equally quick to reject; others requiring more work – and some of those (e.g., ) were also adding links.

I see nothing in the data that makes me think that the newcomer task is disproportionately soaking up WPMATH's time and effort. All in all, if you wanted to reduce the rate of revert-worthy edits, you wouldn't turn off the newcomer tasks; in fact, you might make it mandatory for the first few edits. WhatamIdoing (talk) 02:25, 11 September 2026 (UTC)

The "data" you are examining are not statistically meaningful. There was no actual analysis of the rate of "revert-worthy edits": that would require examining all of them or a large randomized sample, and checking them each carefully, something which has never been tried as far as I can tell. The non-machine-suggested edits that get reverted include vandalism, broken test edits, citation spamming, accidental addition of factual errors (e.g. people try to "correct" a formula that was already correct), grammar garbling, etc., but everyone accepts this time penalty as the cost of having Wikipedia, an encyclopedia "anyone can edit". At least (typically) a human made the decision to make the offending edit, not a robot. There is also no reason whatsoever to believe that mandating "newcomer edits" would help anyone with anything. That sounds like a horrible idea. –jacobolus (t) 03:26, 11 September 2026 (UTC)

Surely, if someone were just to check everything "carefully", the results would be different? WhatamIdoing (talk) 04:45, 11 September 2026 (UTC)

I don't understand what you are trying to say. –jacobolus (t) 05:22, 11 September 2026 (UTC)

I've given you the data that is readily available. That data might not be perfect, but it's what we've got, and there's no particular reason to think that the patrolling that has happened has been biased against your POV. (In fact, since there's been a big discussion at WT:WPMATH and lots of convenient links for patrolling newcomer tasks in this discussion, if the edits were reviewed in a non-random pattern, it's probably biased in your favor.)

You have responded to this data with an assertion that the data is wrong: It wasn't formally randomized. It wasn't done "carefully". Nobody "tried". It reminds me of freshman chemistry labs: Draw the curve, plot the points, and only then take the measurements, because that's the only way to get the answer you want. You (and I, and everyone) know what answer you want. Maybe you should take some measurements now?

Please set a timer and open your watchlist, and tell us how long it took you to review the newcomer tasks you found there. Bring us a few diffs and explain what made them slow and difficult. Above all, please tell us why you're complaining about how time-consuming it is to revert editors doing newcomer tasks, when exactly zero of your last 10 mainspace reverts were editors using newcomer tasks ( ).

From what I'm seeing, you're having a problem with X and blaming it on Y. I'd like to see some data that proves my impression wrong. WhatamIdoing (talk) 06:16, 11 September 2026 (UTC)

My assertion is that your numbers are not relevant, and are not being presented in an accurate way. You are giving numbers for revert rate and then presenting it as a proportion of "revert-worthy edits". But you actually have no way of knowing how many of the edits involved were checked at all, or, if checked, how careful that checking was.

Of ones that I have examined there is a very high error rate. Far higher than I think is acceptable for a bot. If it were like 0.1% errors, and those were borderline cases, then people could probably ignore the edits to pages they care about. But my impression is that it's more like a quarter to a third "revert-worthy" links. But even if it were only 5%, that's dramatically unacceptably high for edits being made by a bot.

I only watch a small fraction of all math articles, so I wouldn't have ever seen most of the 2000+ such edits that you say have been done. The only way mistakes are reverted are if people actually double-check them, but many pages have no active page watchers at all.

Bad wikilinks are harmful to readers. They are confusing. They are misleading. They add visual noise to the text. –jacobolus (t) 06:46, 11 September 2026 (UTC)

• You keep talking about this like a bot is making the edit, when that's not true. A human is making the edit. A human decided to link to Quadratic equation in that article. There is no bot involved, and the software only made a suggestion – to a human, who apparently thought it was appropriate and helpful, even though it doesn't seem that way to you.
• I don't think that accurate links (i.e., those pointing to the correct Wikipedia page) are confusing. What could be "confusing" about a link to Quadratic equation in a sentence that mentions that exact type of equation?
• I don't think that accurate links are misleading. What could be "misleading" about a link to Quadratic equation in a sentence that mentions that exact type of equation?
• I don't think that links add visual noise to the text. I have seen a couple of editors over the years say they are very sensitive to the shift in colors. However, I'm not, and I understand that this is a small minority of people.
• I think that the anti-link POV overlooks the idea that links serve multiple purposes. It's not just – or even primarily – that we should have a link if you need to understand what ____ means to understand the sentence. Links also exist because you might want to navigate to the other article. For this purpose, the relevant question isn't just "Can the reader understand this paragraph if they don't know what this term means?" Instead, there are multiple relevant questions, and one of them is "If you're going down a rabbit hole of Wikipedia articles, might you be interested in looking at this one next?" In technical articles, that can mean linking to simpler articles, especially in the lead, so that someone who's in over their head, or not so interested in this article, has a path back to their preferred depth.

WhatamIdoing (talk) 16:52, 11 September 2026 (UTC)

What could be "confusing" about a link to Quadratic equation in a sentence that mentions that exact type of equation?

Did you actually read the wikilinked article Quadratic equation, or look at the context where it was linked?

Here is how the article starts:

In mathematics, a quadratic equation (from Latin quadratus 'square') is an equation that can be rearranged in standard form as $$ax^{2}+bx+c=0\,,$$ where the variable ⁠$x$⁠ represents an unknown number, and a, b, and c represent known numbers, where a ≠ 0.

The rest of the article continues discussing this type of single-variable quadratic equations, focusing on their solution.

By comparison, the context of the wikilink is:

Descartes' theorem states that the bends of the four circles, $${\begin{aligned}k&={\frac {1}{r}}={\frac {1}{s-c}},&k_{a}&={\frac {1}{r_{a}}}={\frac {1}{s-b}},\\k_{b}&={\frac {1}{r_{b}}}={\frac {1}{s-a}},&k_{c}&=-{\frac {1}{r_{c}}}=-{\frac {1}{s}},\end{aligned}}$$ which are the curvatures of the circles, apart from a minus sign factor for the outer circle, satisfy the quadratic equation $$\left(k+k_{a}+k_{b}+k_{c}\right)^{2}=2\left(k^{2}+k_{a}^{2}+k_{b}^{2}+k_{c}^{2}\right).$$

You will notice that this is not "an equation that can be rearranged in standard form as $\color {PineGreen}ax^{2}+bx+c=0$." So if someone clicks this wikilink, they will get to a page that doesn't even mention the type of equation being discussed at the context where the link is. If someone managed to get to that point in the article without knowing what a quadratic equation was, clicking this link would make them more confused than before they clicked it.

Beyond that, this is a confusing wikilink to put halfway down the page. The entire article Pythagorean triple is full of quadratic equations, from top to bottom. If we wanted to wikilink this term, the reasonable place to put it would be right near the top, and the link should probably be on the word "quadratic" and go to a target like Quadratic function § Bivariate and multivariate cases. –jacobolus (t) 17:33, 11 September 2026 (UTC)

¯\_(ツ)_/¯ I don't find it confusing, and I think that readers will expect the link to take them to the article that it would take them to. WhatamIdoing (talk) 17:42, 11 September 2026 (UTC)

You don't find it confusing to wikilink a technical jargon term, and have the wikilink point to an article that makes no mention of the meaning of the term used in the context where the link was? What do you think is the benefit of such a link? –jacobolus (t) 17:45, 11 September 2026 (UTC)

The overall feeling I get from your comments is that "shrug" basically sums up your attitude toward both our technical articles and the human contributors working on them. –jacobolus (t) 17:48, 11 September 2026 (UTC)

On the specific point of the link in question: I agree with jacobolus that Quadratic equation, an article about a single-variable equation, wasn't an appropriate link for the text in question, which was about a multiple-variable equation. isaacl (talk) 21:28, 11 September 2026 (UTC)

Update: I removed the word "quadratic" from this place in Pythagorean triple. It is not really necessary. –jacobolus (t) 18:27, 11 September 2026 (UTC)

The more I look at this discussion the more I believe that it has turned into a place to complain and rant about this particular feature rather than actually have productive discussion, ie: "should add a link be disabled on math articles?" (appropriate responses being "yes" and "no") Looking at WT:WPMATH, it seems that discussion on the newcomer tasks seems to only involve linking to if and only if, which to me screams out whataboutism. I don't really get jacobolus's point that "(WhatamIdoing)'s numbers are not relevant". From WhatamIdoing's bar chart (comment on 02:25, 11 Sept), edits made by the "add a link" newcomer task only make up a small fraction of total edits, so claiming that the newcomer task is disproportionately soaking up WPMATH's time and effort is... faulty. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 07:33, 11 September 2026 (UTC)

In any case, the WMF team behind it is well aware on this issue and has pushed changes to the system. I suggest that discussion be continued there (I am considering to close this one, and the other one at the WP:VPMISC to centralise discussions). The erroneous links might not show up anymore with the system update! Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 07:47, 11 September 2026 (UTC)

"Disproportionately" to what? I only claimed that the attention and effort required is disproportionate to the minimal benefit this feature has, and that it's frustrating to spend our time on machine-generated edits that have a very high error rate, instead of on the human activity we come here for. –jacobolus (t) 08:59, 11 September 2026 (UTC)

@Hason-LEK-SIN: (and @WhatamIdoing) FYI, I often try to avoid reverting by clicking on the revert button because I think this is not very nice, especially to newcomers: if one of my first edits had been reverted, I think I would have been mortified and would have stopped contributing. So what I usually do, is if I see that that a person has added 3 links, I'll remove two of them manually and try to see if I can keep at least one. That definitely takes more than 30 seconds. Malparti (talk) 15:27, 11 September 2026 (UTC)

@Hason-LEK-SIN I agree that the discussion has drifted a bit from the original question, so let me sum up the situation:

• the best proxy that we have for whatever consensus might exist among math article editors is whatever consensus might exist on WP:WikiProject Mathematics;
• the overall feeling at WP:WikiProject Mathematics seems to be that the link suggestions features is a pain in the neck. This situation may or may not restricted to math articles; but in the case of math articles, denying it would be going against what pretty much every editor of math articles who cares to exchange with other such math editors thinks.
• jacobolus comes here to report this and ask if the feature can be disabled for math articles.
• many people who do not frequently edit math articles jump in to defend the link suggestions feature (and explain to math editors who think it is an extra burden for them with no real benefit that in fact we are wrong to think it is an extra burden).

Now, two questions: (1) do you disagree with my account of things? (2) Ggiven this state of affairs, what should the next step be? Malparti (talk) 15:55, 11 September 2026 (UTC)

I somewhat disagree with it. A quick search through Wikipedia talk:WikiProject Mathematics/Archive/2026 found zero discussions about newcomer tasks or adding links this year. The recent one, which prompted this discussion, was about adding a link to a (single) specific article. It is therefore difficult to sustain a claim that there actually is an "overall feeling" about link suggestions, rather than a one-off incident.

I am also concerned about the problem of misattribution. For example, around the time that Jacobolus started this thread, he reverted this edit for having bad links. That single edit has taken a lot of time and effort for him to untangle (because he's tried to preserve the parts that were good). But it's important to note that edit wasn't made by a newcomer, and it didn't use the newcomer task software. This kind of thing can cause people to be upset about "bad links", without realizing that most of the bad links they're seeing have nothing to do with the newcomer task. Preventing that edit would have saved him a lot of time, but turning off newcomer tasks would not have prevented that edit. The (two) newcomer links he's reverted appear to have taken very little time, especially by comparison to that one. WhatamIdoing (talk) 17:40, 11 September 2026 (UTC)

I am not complaining about the links made in the edit you linked here, which is entirely unrelated to this discussion. I have no problem reverting, fixing, discussing, etc. human-generated edits. They don't cause me to "be upset about bad links"; I have lots of patience for interacting with human editors. I just don't like cleaning up after bots. –jacobolus (t) 17:43, 11 September 2026 (UTC)

I think that the links added to "the number of grains of sand that can be contained in the universe" were bad, and I understand that you did too.

Why do you keep saying that humans using software are "bots"? WhatamIdoing (talk) 18:26, 11 September 2026 (UTC)

Because the links added by the bot suggestions in "newcomer tasks" are categorically different from the type of links that are typically added by humans. The bot that generates these links works by some kind of text search/matching process that does not resemble the process by which humans pick wikilinks to add, and the resulting wikilinks often don't make any sense from a human perspective. When humans add bad wikilinks, I can understand where they are coming from, and if I talk to them I can say something like "I see why you added this, but that doesn't match Wikipedia conventions, go look at MOS:OVERLINK" or whatever. When a bot adds a link via a "newcomer" human proxy, I can't give the bot any meaningful feedback. It's going to continue to make the same type of mistake over and over until reprogrammed. –jacobolus (t) 18:31, 11 September 2026 (UTC)

Which edits are you comparing? What is your sample size? Do you have any data showing that newbies who add links make different/worse ones if they're considering a suggestion vs choosing it themselves? Can't you see why a newbie who added a superfluous link to (e.g.) If and only if might have thought it was appropriate, regardless of what software they were using at the time?

I agree that there is an unfortunate amount of repetition in some of the suggestions. We have an existing system that excludes (e.g.,) the names of common occupations and countries, and it would be nice if we could add locally chosen articles to that list. WhatamIdoing (talk) 01:15, 12 September 2026 (UTC)

I'm just telling you my general impression. I haven't been keeping careful track of every Wikipedia edit I review.

But again, if you want to have some kind of statistically meaningful data, you need to start with the whole population or a large random sample, and then you need to apply some kind of uniform checking process.

If you can come up with a complete list of all of the suggested link edits to math-related articles, and post it somewhere on Wikipedia, someone can perhaps spend a bit of time reviewing a random sample of them. (Frankly someone should probably review all of them; I expect there are hundreds of edits there that should be reverted.)

Can't you see why a newbie who added a superfluous link to (e.g.) If and only if might have thought it was appropriate

Yes, which is why we shouldn't have a bot repeatedly recommending newbies add such links, especially to obscure and technically advanced articles they don't understand. Newbies making such decisions on their own happens rarely, and mostly affects popular articles with many page watchers. –jacobolus (t) 02:45, 12 September 2026 (UTC)

If you know of someone who would like to review all of them, then it'd be easy to produce a list of all such edits to WPMATH-tagged articles during any arbitrarily chosen month(s) or year.

Creating a list of comparable non-newcomer-task edits for comparison would be harder.

Blinding the reviewer to the tags should be feasible with CSS.

But there's no point in doing that if there aren't people willing to spend days re-reviewing the edits. WhatamIdoing (talk) 03:15, 12 September 2026 (UTC)

Can you make a complete list? Not a particular month, but all of them, ever? I don't think a comparison to other edits is going to be that useful, since those involve all sorts of random stuff, including a lot of spam, test edits, and vandalism. –jacobolus (t) 03:37, 12 September 2026 (UTC)

Yes, that can be done via a Quarry query, assuming it doesn't reach the timeout limits. (You might have to break it up, e.g., year by year, to get the whole list without timing out.) See quarry:query/109211: 239 reverted + 6,373 unreverted = 6,612 total. At a reviewing speed of one edit per minute, that's 11 hours' work.

Without a control group, it's not possible to make any comparison to non-newcomer-task edits. The only comparison that would be possible is before/after. If one person reviewed them all, it could be a "How different is my view from ordinary reviewing by the rest of the community?" metric, but it wouldn't tell you anything about whether the newcomer tasks match the community's view. WhatamIdoing (talk) 05:38, 12 September 2026 (UTC)

I just don't like cleaning up after bots.

The "add a link" suggestion may be suggested by the bot, but ultimately, it is the editor who will either accept or reject those suggestions, so I disagree on the fact that those edits were made by a bot. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 22:44, 13 September 2026 (UTC)

@WhatamIdoing Well, I've got news for you: WikiProject Mathematics is not a talk page whose archives you can easily search. It's a community of people who recognize each other's usernames and exchange in various places and forms: article talk pages, user talk pages, thanks on edits, other wikis, and possibly on other platforms than Wikipedia...

When you have a group of friend where everybody agrees that John is kinda weird, it doesn't mean you'll find it written in the group chat. Malparti (talk) 21:31, 11 September 2026 (UTC)

(In fact, most of what I know about the members of Project Mathematics comes from seeing their edits in my watchlist over the past few years, not from talking to them directly) Malparti (talk) 21:45, 11 September 2026 (UTC)

@WhatamIdoing I asked you two very precise questions.

• You answered the first one by saying you "somewhat disagree" with my account of things — more specifically, it seems you disagree with a very specific point. You've explained why, and I've explained why in my opinion your argument isn't convincing. Since then you've not replied. So, to make some sort of progress in that dialogue of the deaf: can we, for now, assume that my account of things is mostly correct, and see where that would take us? If the implications of that assumption warrant it, we can spend some energy double-checking it with a better methodology than Ctrl-Fing the archives of the talk page of WikiProject Mathematics.
• What about my second question? You have not even tried to answer it. My impression is that, so far, you've tried to contest pretty much everything that people in favor of jacobolus' proposal were saying — fine, that's useful. But now assume that most math editors think that in the context of math articles, the "link suggestions" feature is counterproductive. In that hypothetical scenario, what could/should be done?

Malparti (talk) 21:42, 13 September 2026 (UTC)

Here are my answers to your two questions, which were "Now, two questions: (1) do you disagree with my account of things? (2) Ggiven this state of affairs, what should the next step be?":

1) Yes, I disagree in part with your account of things.

2) I suggest that people who want to disable it should provide data demonstrating that:

• There are an unreasonably large number of edits to be reviewed,
• That it takes editors who are interested in math articles an unreasonably large amount of time to review these edits.
• That the WPMATH preference for minimizing links is in line with the broader community's view, which has trended towards greater use of links in recent years (e.g., the community's change to OVERLINK to accept one link per section, rather than one link per whole article).

WhatamIdoing (talk) 22:11, 13 September 2026 (UTC)

There is no such thing as a "WPMATH preference for minimizing links". What Wikipedians working on math articles do generally agree on (with the usual amount of discussion/controversy about borderline examples) is that we should follow the manual of style's recommendation that "words and terms understood by most readers in context are usually not linked". Unfortunately the suggested link bot is not sophisticated enough to follow this part of the MOS. There are plenty of math articles where the same wikilink is included multiple times in different sections, which doesn't cause any problem that can't be worked through via (typically) ordinary editing or (in exceptional cases) discussion. Why should it be up to User:WhatamIdoing in particular to decide what a community of editors considers "reasonable" or not? –jacobolus (t) 22:36, 13 September 2026 (UTC)

I didn't say that it should be up to me to decide what's reasonable. Malparti asked me for my advice on what the next step should be. My advice is that they should provide the kind of non-subjective evidence about the alleged problem that I believe would be convincing to the rest of the community. I suggest this because if (to use my first suggested type of data collection) someone says "Oh, my, there's waaaaay too many of these edits happening to our math articles, so we've got to ban these!", and it's pointed out that in a given week, this so-called "too many" is just one or two per day, out of 33,000 WPMATH-tagged articles – and at a rate significantly lower than the rest of Wikipedia – then the community reaction is not likely to sound sympathetic.

If someone were to conclude that a claim of too many by raw count is impossible to sustain, then it's possible to admit that there are relatively few, but then try to convince people that it's more difficult to review links in math articles than for other subject areas. People generally believe that Math is hard, so they might be more sympathetic to that. But you'd have to build that case, with some evidence.

Finally, the newly created (and non-RFC'd) MOS:MATHLINK may or may not align with the community's view. For that matter, Wikipedia:Make technical articles understandable and Wikipedia:Manual of Style/Linking may be out of date wrt the community's view (e.g., UNDERLINK's "An article is said to be underlinked if unlinked words are needed to aid understanding of the article" doesn't mention anything about the use of links for navigational purposes). But if we're going to use the idea that the tool encourages editors to add links that this new section of the MOS discourages as a reason to change the tool, it seems to me that we should first make sure that the community agrees with the new MOS section. WhatamIdoing (talk) 23:29, 13 September 2026 (UTC)

The new MOS:MATHLINK section, intended to clarify how the existing MOS:OVERLINK guideline can be applied to math articles in particular since their intended audience is often narrower and their context is somewhat different than a generic other article, was created based on community frustration that was, in my opinion, substantially (but not entirely) caused by the suggested link feature. –jacobolus (t) 23:38, 13 September 2026 (UTC)

The new MOS:MATHLINK is a substantive change to a guideline that was decided by a handful of people on a WikiProject's talk page, without even being mentioned on the guideline's own talk page, much less having an RFC, which would be the typical process for a proposal to create a new section that adds 300+ words to a guideline. WhatamIdoing (talk) 23:44, 13 September 2026 (UTC)

If you think there's a problem you are of course also welcome to start a different discussion, e.g. at the MOS:MATH talk page, but the number of people who engage either the content or application of Help:Displaying a formula, Wikipedia:Manual of Style/Mathematics, Wikipedia:Make technical articles understandable, etc. is fairly small and most of those people are (a) watching the pages in question and willing to revert or modify things they disagree with, and (b) notice and engage in discussion at Wikipedia talk:WikiProject Mathematics. I don't think I have ever seen an RFC for a change to MOS:MATH (though a search does turn up one in the talk page, about which unicode symbols should be allowed for roman numerals).

Do you also expect RFCs for changes to MOS/Stringed instrument tunings, MOS/Cue sports, MOS/Romanization of Ukrainian, MOS/Road junction lists, etc.? –jacobolus (t) 01:48, 14 September 2026 (UTC)

I expect RFCs for substantial expansions of all guidelines, because that's the ordinary practice of the community. The whole point of an RFC is that the RFC brings more people to under-watched pages. WhatamIdoing (talk) 17:25, 14 September 2026 (UTC)

Feel free to write up an RFC and list it at what you think an appropriate venue would be. –jacobolus (t) 17:30, 14 September 2026 (UTC)

@WhatamIdoing Am I correct in assuming that the only thing you disagree with in my account of things is the overall sentiment among math editors? If so, given that I assume you obviously understand the interest of making assumption to make progress on a problem: I've explicitly suggested we assume for now that the overall feeling among editors of WikiProject Mathematics is that the link suggestions feature is a pain in the neck — just to see where that would take us in terms of what should be done. If it turns out that something significant should be done, it will then be easy enough to poll WikiProject Mathematics to make sure that assumption is in fact correct; but for now I don't see a need to invest that much energy in stabilizing this point, because I don't recognize your name among regular editors of math articles and therefore wouldn't expect you to have an informed opinion on the overall feeling among math editors.

Regarding the two other items in your reply (namely that before doing anything else, we should demonstrate that there are an unreasonably large number of edits to be reviewed it takes editors who are interested in math articles an unreasonably large amount of time to review these edits), I find the first one no entirely relevant (what's "unreasonably large" anyway, and if they're almost always nonsensical why should we tolerate even a few?); and the second slightly inappropriate: why should you decide what's an appropriate amount of time for me to invest reviewing and reverting these nonsensical edits? Malparti (talk) 22:50, 13 September 2026 (UTC)

• "Unreasonably large" is any amount that the community (NB: not me) will accept as unreasonable.
• "Almost always nonsensical" has been proven wrong. "Usually kept after review by experienced editors" is an accurate statement. "Unwanted more often than some people have patience for" is probably also correct.
• I never said that I should decide how you should spend your time. I am saying that if you want to prevent people from editing "your" articles, then you will have to convince the community (NB: not me) that this tool places an unreasonable burden on whoever reviews these edits. I have made two specific suggestions about the kind of edits that might convince the community of the merits of your complaint. For reference, those two specific suggestions are total volume of edits to review and the difficulty of reviewing them (for reference, it took me an average of 30 seconds per link).

WhatamIdoing (talk) 23:40, 13 September 2026 (UTC)

@WhatamIdoing

• "Unreasonably large" is any amount that the community (NB: not me) will accept as unreasonable. → What community? On Wikipedia, "the community" cannot mean something other than "people who care enough to take part in the discussion".

If you look at this discussion, you'll see that here "the community" mostly consists of two math editors (including one who asks you to temporarily accept that his views on this specific question are reasonably representative of a consensus among math editors); and you.

• These two editors — who routinely review links added in math articles through the link suggestions feature — claim that the current amount is unreasonable.
• You — who do not routinely review such links (and probably have never done so, if I base myself on your claim that it would take you 30s to review a link added to a math article) — claim it is. You have a right to think that and to say it here; but I wonder what knowledge relevant to the situation at hand it is you have that makes you think you have something relevant to say about how "annoying to editors vs useful to the project" the link suggestion features is for math articles. Someone here claimed the feature is useful for military history articles, I wouldn't see myself telling them it's not.

• "Almost always nonsensical" has been proven wrong. → My wording was indeed too strong — I should have said "Almost always useless, sometimes plain wrong". "Usually kept after review by experienced editors" might be true (although, as I've already explained, even when I remove the added links, I typically don't undo these edits so as not to WP:BITE; so I'm not sure how you're quantifying this), but it's irrelevant here. I can re-explain why if needed — but I don't think it is?
• I never said that I should decide how you should spend your time. → Yes, you kinda did, fact. Not directly and not necessarily to me. But you did take a cheap jab at jacobolus on that theme by saying How many hours have you spent trying to convince me that these edits are a burden on your time? How many months' or years' worth of edits could you have reviewed during those hours? And, if I remember correctly, you also wrote some comment (which I won't bother trying to find) that was a cheap jab at the expertise of math editors, because it essentially said "I'd be surprised if so many experts had time to do that"; as a chargé de recherche, it's hard for me not to understand this as saying I should have better things to do (which I have, in fact). Finally your whole rhetoric is build around the idea that "it shouldn't take that long" (implicitly meaning that if it does take me that long, I'm not being efficient, i.e. not using my time optimally).
• I am saying that if you want to prevent people from editing "your" articles, then you will have to convince the community (NB: not me) that this tool places an unreasonable burden on whoever reviews these edits. → First, Don't try to "straw man me" into WP:OWN. I've never said anything that should suggest I think I have some sort of author rights on math articles; and no one here is trying to prevent anyone from editing math articles. The only goal is fix a problem — namely that, in the context of math articles, the link suggestions features is at best useless and at worst counter-productive. Understanding this requires more than just making amateurish data analysis; it requires actually looking at several examples of links being added to articles about subjects one understands; one could also argue that it requires some experience teaching and/or writing math at the level of the article in question (which your I don't find it confusing, and I think that readers will expect the link to take them to the article that it would take them to reply to jacobolus' point about the link to Quadratic equation suggests you don't have).

Second, you say then you will have to convince the community (NB: not me), but exchanging with "the community" is exactly why jacobolus started this discussion; and you'll notice that you are the only person who keeps asking for evidence that link suggestions are annoying to math editors. Apart from you (and jacobolus and I), most people seem to have stopped caring. So pray enlighten me — who exactly in $\mathrm {community} \setminus \{\mathrm {you} \}$ will I have to convince, and where do I start? Should I start a RfC? And how will I know that I've convinced the community, in case the RfC simply dies out because not enough editors care?

Malparti (talk) 02:55, 14 September 2026 (UTC)

I also care. I routinely revert mathematics link suggestions as inane. It is annoying to have to spend time doing so, annoying to have to choose between biting newcomers in this way or allowing the encyclopedia to become in this small way worse, and annoying to have a non-mathematics editor here try to gaslight me that it is somehow not annoying. Three recent examples: , , . —David Eppstein (talk) 06:50, 14 September 2026 (UTC)

David, I've not said that it's not annoying, and I've not said that you shouldn't feel whatever emotions you feel about it. In other words, I'm not gaslighting you.

What I am saying is that if someone wants to convince the community to change something, they've got to convince the community that:

• There is a problem.
• The problem isn't Wikipedia:Ownership of content.
• The problem isn't being blown out of proportion. For example, you've provided diffs to three reverts here. However, those three diffs were spread over 8 days. That particular link may be extremely annoying to you, but it's still just three quick reverts in an entire week. The 33,000+ math articles get more blatant vandalism than that in a week.

@Malparti, maybe you could drop a list in this discussion of, say, the more recent ten edits to a math article that added links. WhatamIdoing (talk) 17:41, 14 September 2026 (UTC)

It seems to me like "the community" is not at issue. The issue is that you WhatamIdoing, personally, think that there isn't a problem, think that the problem is "ownership", and think this is "blown out of proportion".

Nothing about this links feature has been decided by "the community". The bot in question is being run, unilaterally, by the WMF. At best, there is some community feedback that is considered (this type of discussion is an example of that feedback, where "the community" comes to report that the bot is a nuisance and the practical results are bad). The folks responsible at the WMF could stop its operation or reprogram it if they want. –jacobolus (t) 18:08, 14 September 2026 (UTC)

Nothing about this links feature has been decided by the community...if you don't count discussions such as Wikipedia talk:Growth Team features/Archive 9#Should English Wikipedia enable the Suggested Links newcomer task? and repeated announcements on dozens of pages

And, again, it's not a bot. WhatamIdoing (talk) 23:30, 14 September 2026 (UTC)

Yeah, and if you read that discussion, which took place on the obscure "growth team features" talk page, you get a few lukewarm "go ahead and try it" responses, plus a couple of people initially enthusiastic and a couple more who are opposed. For example:

• "it looks like no one opposes the idea of trying it, but that we're really hesitant about turning it on for all new users at once."
• "Anyway there doesn't seem to be much engagement with this topic, so for the purpose of establishing consensus, I'll say Sure, let's turn it on and give it a go. It seems like it should be easy enough to turn it back off if the newcomer links become too high maintenance."
• "terrible idea, lack of wikilinks isnt a serious issue for wikipedia and doesnt require a feature. all this will do is further encourage noob editors to add unneeded wikilinks"
• "I like the idea of experimenting with this, but I also hope this will not be so hardened that we can't possibly ever decide to stop it."
• "In general, this looks like a useful feature. [...] What I don't like about the feature is that it does not seem to be learning anything from our feedback."

Overall, participation was pretty limited, and I don't think the discussion accurately reflects the range or distribution of community feelings either at the time or now. –jacobolus (t) 23:50, 14 September 2026 (UTC)

Sure, there are other sources of irritation (blatant vandalism, citespamming, slop, people seduced by slop...), but that's not a reason to make our burden heavier upon ourselves. Stepwise Continuous Dysfunction (talk) 21:35, 14 September 2026 (UTC)

@WhatamIdoing I don't see why I should have to do this time-consuming work, which in my opinion will prove nothing, just because you somehow think this is important. But I won't give you the pleasure of having a "Ha, gotcha!" moment, so I'll kindly oblige. All of the following examples were added through the link suggestions feature, and in the past week only. And even with such a small sample we can already illustrate various reasons why such links can be bad:

• First, there are the dreaded iff's:

• Post's lattice
• Polygonalization
• Locally connected space
• Separation axiom
• Jensen's inequality

• Then come the links that fall somewhere between "somewhat incorrect" to "plain wrong" — in the past week I didn't notice any "plain wrong" examples, but I did find a few "somewhat incorrect" ones:

• Lindemann–Weierstrass theorem (→ here "more elementary" means "simpler, more direct"; not "non complex-analytical")
• Oblate spheroidal coordinates (→ the link should be {infinitesimal volume element}, not {infinitesimal} + {volume element})

• After this, we have links that are useless because they link a basic concept in an advanced article (the equivalent of linking words such as "and", "person" or "sky" in a non-technical article), or because they come way too late in the article. Those are not intrinsically wrong, but like linking iff, they result in (sometimes comically absurd) overlinking:

• Heyting arithmetic (→ very advanced and specialized article, no point of linking "bijection")
• Gerbe → (very advanced article, no point in link "vector field")
• Phase retrieval (→ advanced and specialized article, no point in linking "iterative method")
• Homothetic center (→ the article is actually not that advanced, but linking "division by zero" after what precedes is useless)
• Parity-check matrix (→ the whole article talks about vector, generator matrices, linear independence, etc; so linking "vector space" at the very end makes no sense)
• Total variation (→ somewhat advanced article; here linking parametric curve isn't useful because it is clear from the context what this refers to, and one actually doesn't need to know anything about parametric curves so this is needlessly distracting)
• Polynomial ring (→ somewhat advanced article, no point in linking "infinite set" that late in the article, especially when notion of infinity has already been used in three different places in the article)

the examples above are relatively clear-cut, but it's not always the case; sometimes the usefulness of the links seems OK but is debatable (e.g, Cauchy's integral formula, Gilbert–Johnson–Keerthi distance algorithm). There are many more such "OKish" links, and distinguishing them from the clearly useless links above takes time and energy.

• Next, the links that are a bit silly, either because they are a bit circular or just link seemingly random words.

• Generalized method of moments (→ random word in article title)
• Symbolic regression (→ common word in event name)
• Shape optimization (→ links to something more general too late; a bit like linking "milk" in the body of "pasteurized milk")

• Of course, you also have the links that would be useful, but are too imprecise to actually be

• Generalized Poincaré conjecture (→ a link here would be very nice, but linking this article is not precise enough to be useful: the link should something like infinite special orthogonal group; because all the ingredients to understand what the infinite special orthogonal group is are indeed in Orthogonal group, but technically it's not discussed in that article, and if you don't already know what SO is you have no chance of understanding from Orthogonal group...)

• Finally, you have links which turn out to be correct "by pure luck" but require a lot of time to be checked:

• Hilbert–Huang transform (→ it's not clear from the Wikipedia article which research paper "Phillips [2003]" is — turns out it is Application of the Hilbert–Huang Transform to the Analysis of Molecular Dynamics Simulations — and even if you know what a conformational change of a molecule is, unless you are in the domain the expression "conformational change in Brownian dynamics" could feel a bit strange and make it look like there is a chance the link to Conformational change is not adequate — here it turns out it is so all good; but I spent maybe 10 minutes checking that).

Again: this is just for the past week; these are that would not have been added if it weren't for the link suggestions feature; and this is not an exhaustive list.

Now, I'm going to repeat myself but this list should not be needed to get the conversation moving forward:

1. when several editors on Topic A come report that something specific to Topic A is hindering their work, to the point that they are ready to spend some time and energy to make it stop: why would your first reaction be to question what they say?
2. the time it takes to review these edits is not the only problem here. As multiple people have kept saying: being forced to choose between biting newcomers and letting articles slowly become worse is unpleasant.

I'm not even mentioning other problems posed by this feature (such as the fact that it is frequently used by WP:SPAs to inflate their edit count and gain legitimacy — on the French Wikipedia I've recently had to waste hours reverting hundreds of edits made by bots using the link suggestions and image suggestions features), because these other disadvantages are not specific to math articles; but of course in the end one should not forget to add them up to the other points listed above. Malparti (talk) 22:12, 14 September 2026 (UTC)

Thanks Malparti. Ping @KStoller-WMF. I think it's important that the folks from the WMF see the above post. 22:45, 14 September 2026 (UTC)

@WhatamIdoing Your lack of response to the above post is very conspicuous. I'm going to assume that you are conceding the point here. So now, can you help us get this turned off for math articles? –jacobolus (t) 17:14, 18 September 2026 (UTC)

It's on my list, but I haven't had a larger chunk of time to dedicate to reviewing it during the last three days. I want to be able to respond in greater detail than "Thanks for showing that most of the links are accurate" and "Did anyone add links to any math articles last week that didn't use the newcomer task?" WhatamIdoing (talk) 18:57, 18 September 2026 (UTC)

@David Eppstein A previous post said:

I believe Category:Mathematics could be added to the exclusion list for Add a Link task.[...] An English Wikipedia admin can update the "Articles containing categories defined here will not be shown to users as tasks for this task type" field via Special:CommunityConfiguration/GrowthSuggestedEdits if it seems like there is community agreement that link suggestions on Mathematics articles are particularly problematic.

Is that something you can do? –jacobolus (t) 22:49, 14 September 2026 (UTC)

There are only 8 pages in that category, and there probably shouldn't be any. WhatamIdoing (talk) 23:11, 14 September 2026 (UTC)

If there is consensus for this change, I am capable of doing it. I am not convinced this thread has established that consensus yet. Currently the only categories listed there are Category:Candidates for speedy deletion and Category:Articles for deletion; it does not list any content-based categories. —David Eppstein (talk) 23:12, 14 September 2026 (UTC)

I suspect that the cat restriction is limited to the cat itself, and none of the subcats. I wonder if the template {{math}} would work better. That would pick up 12,000 articles, including 1500 articles containing the words "if and only if" (about 20% of all articles containing that phrase). WhatamIdoing (talk) 00:16, 15 September 2026 (UTC)

RfC: Name of the "Discussion" button

Fifteen years ago, we decided to change the label of the "Discussion" button to "Talk". Should we change it back to the default label? 16:03, 3 September 2026 (UTC)

Current label

Default label

Survey (button name)

Yes. And other language Wikis usually say discussion, eg Italian, French etc. Yesterday, all my dreams... (talk) 16:07, 3 September 2026 (UTC)

Yes because:

• More and more of our readers are assuming that the "Talk!" button on our articles takes them to an AI chatbot.
• As was explained below, users are mistaking it for text-to-speech. Most online news articles nowadays have a button that reads the article aloud for them.
• Most people reading foreign Wikipedias probably have a decent grasp of the language, but enwiki is the largest and typically attracts more non-native speakers. "Talk!" is more recognizable than "discussion" and it's also an imperative verb.
• "Talk" is the more opaque of the two terms for newcomers, which further helps to hide its intended function, which is to "discuss" how to improve the article itself, not to "talk" about the topic of the article. "Talk!" also has a broader meaning and can be used in context where "discussion" cannot.

FaviFake (talk) 20:56, 3 September 2026 (UTC)

I have already seen these reasons, no need to ping in an edit summary please. --ABx11 (she/they) 00:44, 4 September 2026 (UTC)

If it moves the needle even a teensy tiny bit, please god yes. I feel like I am the only person monitoring these, it is one of the few places on Wikipedia that actually DOES have a deadline (since you have to remove it before the archive bot gets it, otherwise you aren't allowed to clean it up anymore, joining the nearly 6,000 instances of enshrined untouchable vandalism), and I am so, so, so behind on it. Gnomingstuff (talk) 20:58, 3 September 2026 (UTC)

@Gnomingstuff: re: "you aren't allowed to clean it up anymore" - says who? Where and when was that decided, and was there broad participation? A rule that requires vandalism to be enshrined sounds like a really bad rule and one that should be reconsidered. HierophantOfOmens (talk) 07:12, 5 September 2026 (UTC)

Please do not ping me to a discussion I am obviously aware of given that I commented in it 5 minutes ago. If I want to look at a discussion, then I will do so on my own time. If I am not currently looking at a discussion, then I'm probably doing something else that I would prefer to not be interrupted during. In this case, it was doing the cleanup this whole thread is about, and now I am not doing it, all thanks to your ping.

The RfC was here. Gnomingstuff (talk) 07:15, 5 September 2026 (UTC)

I don't see how one interprets "There is no consensus one way or the other regarding editing talk page archives for any other reason, such as removing a nonsense post that was not reverted prior to archiving ... This close should not be construed as limiting removal of content from archives for other policy-based reasons, such as the legitimate use of oversight or revision deletion, nor should it be construed as affecting the reversion of vandalism to archives, which I don't believe was ever really in question here" (from the RfC close) as meaning "you aren't allowed to clean it up anymore", or that there is any such thing as "enshrined untouchable vandalism". Levivich (talk) 07:24, 5 September 2026 (UTC)

The RFC's closing statement begins: "There is consensus that talk page archives can be edited to remove material that breaches policy, such as copyvio, libel, and serious personal attacks..."

The first link in "the nearly 6,000 instances of enshrined untouchable vandalism" adds a racial slur to someone else's comment. That's obviously something that should be fixed (if it hasn't already been).

Mostly, when I see this kind of comment, though, I'm reminded of User:Betacommand, whom we tasked with tagging WP:NFCC violations. Then we left him to deal with the social fallout all on his own, and it turns out that having the technical skills to find and tag various files did not automatically give someone an endless supply of patience and kindness to people who felt entitled to ignore the legal policies, or at least to yell at someone before making a show of how they're grudgingly complying with these completely unnecessary, overly picky rules. It did not end well. I am concerned that we are doing the same thing to our AI defense folks. WhatamIdoing (talk) 20:35, 5 September 2026 (UTC)

You know, having learned to dabble (a tiny bit) in the site API... I bet you USD $1 that list of 6000 is bigger. — Very Polite Person (talk/contribs) 14:52, 5 September 2026 (UTC)

Yes to simply renaming the visual presentation of the clickable buttons from "Talk" to "Discussion". Very good idea. We have no justification, need or reason to be the weird outlier versus other wikis, and per other arguments here. — Very Polite Person (talk/contribs) 22:41, 3 September 2026 (UTC)

No per Maddy from Celeste and Very Polite Person below. Firstly there is no evidence that this will solve the problem it attempts to - everybody who thinks "talk" means "discuss this article with a chatbot" will think "discuss" means "discuss this article with a chatbot" so we gain nothing. Secondly the 2011 change was made for a good reason and that reason still exists, we will still refer to the page as a talk page and that will still confuse new editors who cannot find a button for a "talk" page, so we will lose significantly here - possibly even more than we were doing in 2011 as we are facing a bigger editor recruitment problem now than we were then. Thryduulf (talk) 22:50, 3 September 2026 (UTC)

No per ... Very Polite Person below.

I'm confused. Very Polite Person has always supported this proposal from the start, saying it's a very good idea. You're the first person to oppose. FaviFake (talk) 22:59, 3 September 2026 (UTC)

That's because I misattributed a comment, my opposition is per Anomie not VPP (my apologies to both). Thryduulf (talk) 23:05, 3 September 2026 (UTC)

How dare you, etc. All good. — Very Polite Person (talk/contribs) 00:16, 4 September 2026 (UTC)

• No - Using a different word (“Discussion”) on the tab we click to reach the article’s TALK page makes no sense to me. I would expect to click on a tab reading “Talk” to reach the article’s TALK page. Blueboar (talk) 23:26, 3 September 2026 (UTC)

  Which, incidentally, would also be the button you'd expect to click in order to talk with other users (or with a chatbot) about the topic, or to have the article automatically read aloud to you... FaviFake (talk) 23:34, 3 September 2026 (UTC)

  If all the other equivalent encyclopedias apparently use their native tongue "Discuss", why wouldn't we want to be doing it how the others are? — Very Polite Person (talk/contribs) 00:18, 4 September 2026 (UTC)

• Yes per Favi and Gnoming, "discussion" is much more intuitive, and we should be targeting readers' and newbies' experiences here. One of the faults of our decision-making process is that only experienced editors participate and prioritise their own experience. Kowal2701 (talk, contribs) 00:14, 4 September 2026 (UTC)

  The argument in favor of Talk, both in 2011 and now, is that "Talk" is more intuitive since it's the name of the namespace, and it's what everyone calls it, and that it would be better for readers and newbies to therefore have the button called "Talk." They're not prioritizing their own experience, they're targeting readers' and newbies' experiences.

  So does anyone on either side of this debate have any actual data to present, or are we just voting based on our own personal experience/opinion of what is less confusing for others? Levivich (talk) 01:54, 4 September 2026 (UTC)

  True, and it's the latter, no idea if there's a way to request data collection via survey (there should be) Kowal2701 (talk, contribs) 02:42, 4 September 2026 (UTC)

  Do the however many tens of thousands of prompt junk posts I have reverted count as data? The neverending deluge of stuff like this day after day after day, hundreds per week, thousands per month. Gnomingstuff (talk) 06:58, 5 September 2026 (UTC)

  No. I'm not talking about data that people misuse talk pages, anyone that's read some knows that. I'm talking about data about the button label. That edit wouldn't be prevented if the button was called "Discussion." A survey, as Kowal suggested above, and A/B testing, as CMD suggested in the discussion section below, could bring useful data, though. Levivich (talk) 07:12, 5 September 2026 (UTC)

• (edit conflict) Weak no because while talk might suggest that the Talk page is for discussing the subject of articles, discussion isn't better in this regard. Of course, if the proposer has a better idea or wishes to clarify, I am open to changing this vote. --ABx11 (she/they) 00:17, 4 September 2026 (UTC) Hard no. The only reason that has been presented to me via a ping isn't sufficient as I had already seen these reasons. And after seeing the arguments against it presented elsewhere, I'm convinced that this might cause confusion, and frankly, I'm not even convinced that it even helps in this regard. --ABx11 (she/they) 21:22, 5 September 2026 (UTC)
• No per Maddy from Celeste BilledMammal (talk) 00:19, 4 September 2026 (UTC)
• No per Blueboar, Thryduulf, et al, and per the 2011 discussion that lead to the change to "Talk" in the first place. The conclusion drawn then remains very much valid. A tab leading to the Talk: namespace should be labeled "Talk".

Plus I think the idea that somehow "Talk" leads people to think it's a chatbot and "Discussion" will not is an unfounded conclusion. That really would need some sort of actual supporting evidence besides anecdotes, because I haven't seen any such activity. (I have seen many cases where none-too-bright people for some idea think it's a page for contacting the subject of the article, but that can be dismissed as clear lack of competency.)

Finally, what other language Wikipedias do is irrelevant. Each language Wikipedia is a separate entity. Yes, a lot of them use the cognate of "discussion" for that tab. But they also use that cognate for the namespace for their discussion pages. Which circles back to the whole reason the English Wikipedia tab is "Talk". It matches the namespace. To change the tab and not the namespace makes little sense. Either change both or change neither. Changing only one creates unnecessary confusion. oknazevad (talk) 00:41, 4 September 2026 (UTC)

Furthermore, Talk and Discuss are synonyms, so I'm not seeing the tangible benefit. --ABx11 (she/they) 00:44, 4 September 2026 (UTC)

• No Way, way too many references to "the talk page" litter the database and help guides. This would make "the talk page" harder to find and I don't see the value. * Pppery * _{it has begun...} 03:11, 4 September 2026 (UTC)
• No Fiddling with labels won't help people who are unfamiliar with Wikipedia. Having a discussion button that leads to talk is confusing (absurd, actually). Discussion invites commentary even more than talk. Johnuniq (talk) 06:17, 4 September 2026 (UTC)
• No (summoned by bot) Fiddling with labels won't help people who are unfamiliar with Wikipedia. Having a discussion button that leads to talk is confusing To the extent that there may be a problem (not sure of that), this isn't going to provide an answer IMO. All websites have their particularities and at least WP doesn't periodically reframe its appearance for trivial reasons. Pincrete (talk) 07:50, 4 September 2026 (UTC)
• (summoned by bot) No per oknazevad and Pppery. As others have said, what other Wikipedias do is irrelevant, it would lead to more confusion from newbies, and it wouldn't decrease confusion with AI chatbots. Kovcszaln6 (talk) 10:19, 4 September 2026 (UTC)
• Yes "Discussion" is a noun, while "talk" is a verb. Velocifyer (talk) 17:27, 4 September 2026 (UTC)

  From the Merriam Webster dictionary: "Talk: noun ... 4: a formal discussion, negotiation, or exchange of views —often plural" - Donald Albury 19:02, 4 September 2026 (UTC)

  They probably meant to say that "Talk!" can also be interpreted as a verb while "Discussion" can only be interpreted as a noun. FaviFake (talk) 19:40, 4 September 2026 (UTC)

• No Perhaps a bit counterintuitively, contra some of what is said above, "discussion" would lead new people to think it's a place for discussing the article in general i.e. a forum. Jahaza (talk) 20:57, 4 September 2026 (UTC)
• No I do not see a net positive from making the change. - Donald Albury 23:23, 4 September 2026 (UTC)

  I consider "Talk page" to also be acceptable. Donald Albury 18:55, 13 September 2026 (UTC)

• Yes, as a small but hopefully meaningful step in the right direction. I would also be open to changing this for a month or so, and seeing if there was any meaningful difference. Axolitl (talk | contribs) 16:22, 5 September 2026 (UTC)

  I would also support Talk page. Axolitl (talk | contribs) 04:09, 13 September 2026 (UTC)

• No per Thryduulf. —Kusma (talk) 20:49, 5 September 2026 (UTC)
• No. Ain't broke, so don't "fix" it. Seraphimblade ^{Talk to me} 01:33, 8 September 2026 (UTC)
• Talk page. (see ExtantRotations' comment below) I think this is unambiguous and less likely to be confused with LLM slop chatting. I doubt this option will reach consensus, but "yes" and "no" are both bad choices as they have too many drawbacks. --not-cheesewhisk3rs ≽^•⩊•^≼ ∫ (pester) 17:02, 8 September 2026 (UTC)
• No to "Discussion", Yes to "Talk page" per ExtantRotations. --Ahecht (TALK
  PAGE) 18:38, 8 September 2026 (UTC)
• Sure we can always change back again. 15 years seems an appropriate time to try something new. —TheDJ (talk • contribs) 22:22, 8 September 2026 (UTC)
• No. I don't mind causing confusion/alarm/despondency with a change if there's a clear benefit, but here I think it's just speculation to say that one of these words is better than the other. This would create a lot of fussing and complaining and I can't believe it would be worth it. Mike Christie (talk - contribs - library) 23:57, 8 September 2026 (UTC)
• No. I think that the term Talk describes the intended purpose adequately. GrinningIodize (you just lost the game) (talk) 22:35, 9 September 2026 (UTC)
• This is oddly specific, but I would permanently change the button to "Discussion" for only ChatGPT and the other articles on the AI chatbots. Why are people always Googling ChatGPT when they could use Google AI mode? For the other pages, lean yes as a trial. EDIT 02:41: Actually, yes, per FaviFake, and additionally because that's what Commons does as well. Also, I believe it should be made clear that only the button's name is changed, not the namespace. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 02:38, 11 September 2026 (UTC)

  There are way more of these than one would think, I've been requesting semi-protection on the most common offenders but new ones keep popping up

  worst of the lot are the ones on deepfakes because the prompts are specifically people asking for revenge porn Gnomingstuff (talk) 21:09, 13 September 2026 (UTC)

• Oppose change to "Discussion". As User:Johnuniq said, the word "discussion" invites even more commentary than "talk", and it'll make it sound for new users like the comments section from a journalism site. Concerns of "talk" implying a chatbot is IMHO weird, because... how? At most, a new user would reasonably mistake it for a recommendations comment thread or help desk. Plus, {{Talk header}} exists and that cancels out most of the concerns with new users not understanding the purpose of talk page. Too many things in the encyclopedia says "talk page" and making a mandate to change that would be time-consuming and we don't need to fix what is not broken. I would support change to "Talk page" as per User:not-cheesewhisk3rs. In solidarity, Brynn Who Likes Editing | talk w/ me! 07:58, 11 September 2026 (UTC)

  Concerns of "talk" implying a chatbot is IMHO weird, because... how?

  I don't know what to say here, if people who actually do talk page patrol are telling you that this is a thing, then it seems at least worth considering that they know what they are talking about

  see also the edit histories of basically any even tangentially AI-related page, many of which (e.g. Talk:Google Gemini) have banners added by people who could tell you that this is a thing Gnomingstuff (talk) 21:11, 13 September 2026 (UTC)

• No The "Article" button takes the user to the page in articlespace. The "Talk" button takes the user to the page in talkspace. Giving more names to this space would likely just add to the confusion, and "Discussion" doesn't make the purpose much clearer either. Would likely support "Talk page" though. FlammablePizza (talk) 01:07, 13 September 2026 (UTC)

  Article and talk aren't the only namespaces though. On Wikipedia pages it says "Project page", in MediaWiki it says "Interface page", User says "User page" etc. Axolitl (talk | contribs) 04:08, 13 September 2026 (UTC)

  That being said, "Talk page" is a valid option. Axolitl (talk | contribs) 04:09, 13 September 2026 (UTC)

• No to "Discussion", Yes to "Talk page". I'm with others here that changing it to 'Discussion' would be unnecessarily confusing re: the discrepancy between that term and the namespace, as well as with all the existing references to talk pages that are all over essays, help pages etc. Changing it to "Talk page" seems to elegantly solve all of the issues that people have with "Talk" without running into the problems with "Discussion". Athanelar (talk) 00:47, 14 September 2026 (UTC)
• No with caveats as per Blueboar and the 2011 discussion, with my caveat being that we try and make sure all talk pages have the this is a talk page notice for new users. I don’t know if there’s a tool for speeding that process up, though I’m sure it’d be easy to implement in tools such as rater, though that is a separate discussion.Edlectro28 (talk) 15:42, 22 September 2026 (UTC)

  Unfortunately that's controversial and we would need a different RfC for that, but, luckily for you, I'm already working on it! Please feel free to comment at WT:RFC § Add the Talk header template to existing article talk pages before the RfC goes live. FaviFake (talk) 15:48, 22 September 2026 (UTC)

  I got carried away! Anyway, my no still stands, and that’s great timing and thanks for letting me know. I’ll be sure to say my piece over there. Edlectro28 (talk) 15:52, 22 September 2026 (UTC)

  Perfect! (To clarify, the actual RfC will happen on this page, that's just the workshop discussion.) FaviFake (talk) 16:06, 22 September 2026 (UTC)

  Thanks Edlectro28 (talk) 16:08, 22 September 2026 (UTC)

  Change to "Talk page" as this can solve the problems mentioned in this RFC and the one 15 years ago. Szxq (talk) 04:08, 23 September 2026 (UTC)

Discussion (button name)

Presumably, "Talk:Foo" would automatically be created (or maintained) as a redirect to a given "Discussion:Foo"? BD2412 T 16:53, 3 September 2026 (UTC)

No, only the label of the button would be changed, not the name of the page itself. The pagename of the talk pages have always started with "Talk:", even before we changed the button label to "Talk". FaviFake (talk) 17:01, 3 September 2026 (UTC)

Wait, so this proposal is JUST changing the visual presentation of the buttons on web and mobile versions of the page from "Talk" to "Discussion"...

...and that's it? If so can you make the RFC a teeny bit clearer that it's a MUCH smaller but impactful change, and as editors there is no actual impact? — Very Polite Person (talk/contribs) 22:40, 3 September 2026 (UTC)

Yes, exactly!

 Done, and i also added a side-by-side image comparison for clarity FaviFake (talk) 22:52, 3 September 2026 (UTC)

Why change just the label and not also the namespace? Levivich (talk) 23:07, 3 September 2026 (UTC)

The default namespace is fine. When editors open a talk page, they're almost always greeted by a {{talk page banner}} that immediately tells them in a bolded font that "This is the talk page for discussing improvements to the [Weather] article. This is not a forum for general discussion of the subject of the article." We couldn't ask for a better notice.

The button, on the other hand, is completely ambiguous: it just says "Talk!", which is website design means "talk with our chatbot!" or "chat with the community!". There's a reason "Discussion" has always been the default. Plus, changing the namespace would be a nightmare, while changing the label of a button literally takes 1 second. FaviFake (talk) 23:20, 3 September 2026 (UTC)

It probably isn’t just chatbots — it really started to accelerate in early 2022 (before ChatGPT and derivatives) and some comments show indicators that the users are mistaking it for text-to-speech, Siri, etc. No I don’t have any diffs handy but I have personally dealt with tens of thousands of these Gnomingstuff (talk) 23:38, 3 September 2026 (UTC)

I remember a phase from around then of IP users creating talk page sections containing a short phrase repeated in the heading and body (eg. this where a user posts to Talk:Plain with a heading Structural plain and a comment of structural plain). They read like someone trying to search for specific information or navigate to a different article, where they were re-entering the term a second time as a comment in order to make the "Add topic" submit button become clickable. I think some kind of filter was added to stop very short new-thread comments from being posted?

The bigger issue still seems to me to be the new thread talk box itself, with its deeply cryptic request that the user enter not a comment but a "Description". Belbury (talk) 17:28, 8 September 2026 (UTC)

If there's a filter, I don't know about it and it's not working very well. Here's an example from 2 hours ago. (The person they pinged has only ever made one edit, in 2021, to their userpage, so there's virtually no way this is a legitimate attempt to ping them.) I've tried to get a filter put into place, but no one ever did it.

The reason you see this stuff less often now is that I have been going through every single edit made to talk pages, every single day (except when I get behind), since 2025. Gnomingstuff (talk) 07:54, 19 September 2026 (UTC)

Oh, this is the talk page? I wanted the discussion page. I must have clicked the wrong thing. ~2026-47946-04 (talk) 07:21, 4 September 2026 (UTC)

Given that this plan is only to change the display text on the icon, why not change the text to Talk Page instead simply Talk? It removes the ambiguity of being an LLM and is still fewer characters than the current proposal. Plus it remains in line with the existing set up. ExtantRotations (talk) 16:40, 5 September 2026 (UTC)

I think one word, without "page", would be more friendly to newbies and readers. "Discussion" and "talk" are words used for discourse surrounding something, so its purpose as a button seems more obvious. If you were new and saw "talk page", I don't think it would be clear that it is for discussion of the article. Axolitl (talk | contribs) 17:36, 5 September 2026 (UTC)

You know what, I take back what I said. I would not be opposed to adding "page". Axolitl (talk | contribs) 20:37, 5 September 2026 (UTC)

Contra Axolitl, I think this is a very sensible suggestion with few downsides. "Talk page" is no more or less obviously for discussing an article than either "talk" or "discussion", it matches what we tell new editors to look for and the elision of "page" in casual discourse is hardly unintuitive. Thryduulf (talk) 17:39, 5 September 2026 (UTC)

This is a good idea, and will avoid rewritting years of documentation calling it the “talk page”. Velocifyer (talk) 19:15, 5 September 2026 (UTC)

With the exception of the first, the labels are supposed to be verbs. It's an article: talk about it, edit it, view its history, etc. WhatamIdoing (talk) 20:39, 5 September 2026 (UTC)

I would support Talk page as well. --Ahecht (TALK
PAGE) 18:36, 8 September 2026 (UTC)

Re some of the comments above, is there any reason to think clueless people who think "talk" means talking to an AI chatbot wouldn't think the same for "discussion"? Anomie⚔ 22:09, 3 September 2026 (UTC)

Language barrier possibly, “talk” is a more recognizable word than “discussion” and it’s also an imperative verb

to be clear I don’t expect this to have an earth shattering effect but literally any tiny improvement would help Gnomingstuff (talk) 23:40, 3 September 2026 (UTC)

Would be curious though to hear whether other wikis have this problem to the same extent (proportionate to their size obviously) Gnomingstuff (talk) 23:41, 3 September 2026 (UTC)

I don't think other wikis would have this issue because the default is not "Talk" and, as Yesterday suggested above, most other wikis haven't changed the default. I'm an admin on a smaller wiki myself and we've never noticed editors coming to the talk page to have the article read aloud to them or chat with a bot. I'd guess this is partially due to the fact that we've always kept the default label.

But of course I'd love to know if this problem affects other larger wikis. FaviFake (talk) 23:51, 3 September 2026 (UTC)

Talk:Google sees this kind of disruption about once every few days. de:Diskussion:Google about every few months. Google gets 11 times as many pageviews as de:Google. So they're on the same order of magnitude, but maybe we get a bit more problems per pageview than them? There's so many possible confounding factors though that I wouldn't really put any weight on something like this. For one, most people reading dewiki probably have a decent grasp of German, whereas enwiki is typically a top search result for everyone, so it's to be expected that we get a lot more people who can't really understand what these things mean. ;; Maddy from Celeste (WAVEDASH) 00:14, 4 September 2026 (UTC)

We should remember that other language Wikis have distinct cultures. The French Wiki crowd often know each other surprisingly well, almost like the usual customers at a cafe in Paris. The Italian Wiki has far fewer vagabond editors and apart from hot topics like fascism is rather stable. The Spanish Wiki is somewhat similar. The German Wiki is highly controlled, more than one would expect and most users know the rules. But they all use discussion as the button text. Yesterday, all my dreams... (talk) 05:06, 4 September 2026 (UTC)

I would encourage people to go and have a look at that 2011 discussion. The consensus there was that having the link text differ from the namespace's name was confusing to newcomers, who were being told to go to the "talk page" but couldn't find a "talk" link. I don't see why this wouldn't still apply. ;; Maddy from Celeste (WAVEDASH) 22:43, 3 September 2026 (UTC)

• This sounds like something that could genuinely, and without significant collateral, be A/B tested. CMD (talk) 03:04, 4 September 2026 (UTC)

  At the very least the placeholder text could be A/B tested (although it would be nice to get a heads up). The reason so much of this stuff has lines like "English", "History", "Social science"/"SST", "MAPEH" etc. is because people who want to cheat on their homework see "Subject" and assume it is a chatbot asking them what the school subject is. Gnomingstuff (talk) 07:02, 5 September 2026 (UTC)

• A point I forgot to make is that this is a very frequent vector for posts requiring oversight. People will post their actual phone numbers, addresses, bank account numbers, you name it, because they think it's ChatGPT. So if we can discourage that even a little, it's a clear net positive. Gnomingstuff (talk) 06:59, 5 September 2026 (UTC)

  Speaking as an Oversighter, I see no evidence that prompts for Chat GPT are a common reason for posts being oversighted. I've looked at the most recent 20 suppressions on talk pages and only 1 of them seems likely to be an LLM prompt, people have been posting phone numbers, bank accounts, etc to Wikipedia since long before LLMs were a thing (subjectively I don't think it's increased significantly either). I also see no evidence that suggests changing the label of the page will reduce the incidence of people doing this. Thryduulf (talk) 10:57, 5 September 2026 (UTC)

• Wait, who added this RfC? The signature appears malformed. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 08:12, 5 September 2026 (UTC)

  It's normal to end the statement of the RfC question with just a timestamp, since it isn't supposed to express anyone's opinion. ;; Maddy from Celeste (WAVEDASH) 08:16, 5 September 2026 (UTC)

  I see. Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 08:21, 5 September 2026 (UTC)

  This is officially permitted. About 10% of RFCs use that style. A lot of editors never even notice. WhatamIdoing (talk) 20:44, 5 September 2026 (UTC)

Is there empirical evidence supporting either the contention that the current "Talk" label is confusing for some new editors or the contention that changing the label to "Discussion" would be confusing? I'm not comfortable making assumptions and this is something that is testable. ElKevbo (talk) 16:48, 7 September 2026 (UTC)

There wasn't any evidence when we first decided to change it and there still isn't any, afaik. FaviFake (talk) 17:02, 7 September 2026 (UTC)

"We never knew what we were doing before, why start now?" -- Wikipedia, in a nutshell. (Somehow, it actually works.) Levivich (talk) 17:57, 7 September 2026 (UTC)

There was plenty of anecdotal evidence from people who dealt with new editors, e.g. , , . The argument in favour of "discussion" doesn't have even that. Thryduulf (talk) 18:11, 7 September 2026 (UTC)

New proposed experiment to let readers know about “Google preferred sources”

Hi all,

Maryana from the WMF Product and Technology department here (Future Audiences team), coming to let you know about a new Google feature and proposing a short experiment to get data about it. Tldr: the proposal here is about running a temporary (2-week) notice asking readers who come from Google to set Wikipedia as a “preferred source” in Google. We would see if this causes people who use Google to visit Wikipedia more. More info below:

What is Google Preferred Sources and how does it work?

Sketch of what Google search results look like with and without Wikipedia selected as a preferred source

This is a new feature that lets signed in Google users select “preferred sources” that they want to be used as citations in AI overviews on the Google search results page. Any website can be added as a “preferred source,” and that website will then a) appear as the first cited result in an AI overview and b) get a “preferred” icon added to it to make it more prominent/entice clicking on to read more. Google claims that preferred sources get twice as many clicks as other sources cited in AI overviews. If you have a Google account, you can test this out right now by adding Wikipedia (or any other source you like) as a preferred source and seeing how that changes what you see in search results. It’s important to note that no other search engines offer something like this at the moment. For now this is just a Google thing.

Why does this matter for Wikipedia?

Google is the number one way most people get to Wikipedia today, by a huge margin. About 75% of the billions of reading sessions on Wikipedia start on a search engine, and 90% of that comes from Google. What’s quite scary is that the absolute number of reading sessions that come from Google is shrinking every month. There were about 138 million fewer clickthroughs to Wikipedia from Google in the last week of August/first week of September this year than last year (that's a 17% year-over-year decline). Fewer clickthroughs = fewer readers and editors to support and maintain Wikipedia.

With things changing like this, I think this is a time that we need to try things to see what works. I assume everyone here believes that Wikipedia is a great place to learn, so we want people to keep seeing and clicking on Wikipedia links in their Google search results to come here. And I think if a search engine is offering its users a way to see Wikipedia links more often and prominently in search, this is something we want to make known to Wikipedia readers. Would this feature actually make a difference in terms of getting readers to come to Wikipedia more? I don’t know, but rather than speculate, I’d like to run an experiment to find out :)

What's being proposed?

How this notice could look. It would only show up when someone comes from Google to Wikipedia on mobile and won’t show up on subsequent pages. Swiping away or opening “learn more” and dismissing will make it not show up again. If a user clicks through, they’ll go to their Google source preferences page with wikipedia.org prefilled, and they can then choose to save or not.

I’d like to run an experiment where we show a small notice about Google preferred sources to logged out mobile web Wikipedia readers coming from Google, for two weeks. I’m interested to see if this changes how often those readers come back to Wikipedia in comparison to readers coming from Google who didn’t see the notice (i.e “reader retention”). People who see this notice and aren’t interested will be able to dismiss it, and after two weeks, this notice will be turned off and I will share the results here with all of you. I’d like to run the test fairly soon (ideally in late September/early October) so we don't confuse readers seeing fundraising banners in the coming months.

I just want to be extra clear that this is not a situation where an experiment would be an irreversible precursor to full rollout. The notice will be turned off when the test is complete and then we can discuss results together to figure out how to proceed.

Concerns/Transparency

On the other hand, what might be some reasons that would give us pause? I can think of a few reasons, including:

• It could look like advertising for or elevation of Google (not my goal! And this is why we’d restrict the notice only to readers coming from Google, who are already using it – as well as because no other search engine offers something like this).
• It could look like we’re encouraging people to create personalized information silos (it's pretty clear that people already are doing this, and probably better if Wikipedia is in there, but we could debate this).

This is just to say that while other media sites are jumping on the bandwagon of heavily promoting this feature, I don’t think we’re there yet, or that we definitely should get there, but this feature is an option for us and I’d like us to have data.

Finally, I’m also planning to approach a few other language Wikipedia communities to propose a test, as well, but of course English is the biggest language we have and our fastest way to get statistically significant results.

What are your thoughts and concerns? Are there any other risks we should consider before proceeding with testing? Is there a strong reason not to run an experiment like this to get this data before the end of this year? Maryana Pinchuk (WMF) (talk) 15:07, 9 September 2026 (UTC)

My two cents here: I heavily prefer the one-click(tap) preferences setting over two clicks(taps). The statistics on source-visual editor switching (which requires even more clicks) show that the longer it takes to get to something, the less likely it is to be used. The Google source preferences page is already fairly explanatory on its own. -- Reconrabbit (talk) 16:00, 9 September 2026 (UTC)

From a security perspective, selecting a link on one site shouldn't be able to trigger a change on another site. Also, if you're counting on users reading the explanation on the Google soource preferences page, then there needs to be one click to reach that page, and then another page interaction to configure the source preference and submit it. isaacl (talk) 17:53, 9 September 2026 (UTC)

Yep, exactly this – while it's definitely much more user-friendly from an end-user perspective, behind the hood, a one-click interaction is one that sends data on behalf of the user to a third-party site, and that's something that would require thorough security review (which is outside the scope of a quick lightweight experiment like the one I'm proposing). The Google source preference settings page (not the marketing page about the feature, but where you actually go to select preferred sources) is also very light on information about the feature and what it does, so my worry about just showing a notice that takes readers to that page (not interactive, but without any intermediate explanation of the feature), is that people who click through won't have enough context on what they're looking at to know what to do there or why they should do it. I guess we could link to the marketing page (which does have a link to the settings page) but that still means we'll lose some % of people who won't click through to source settings or know how to select Wikipedia. Maryana Pinchuk (WMF) (talk) 18:13, 9 September 2026 (UTC)

Following The Guardian's example, having the explanation frontloaded on the website before you get sent to a preferences page is preferable; with that site, though, you'd have to intentionally click on the article about preferences before seeing the "Click here" that autofills theguardian.com. -- Reconrabbit (talk) 18:39, 9 September 2026 (UTC)

Sounds like a good idea, a lot of people might not have heard of this (for instance this was how I learned it existed). Gnomingstuff (talk) 02:00, 10 September 2026 (UTC)

My main concern is that Google AI search results are absolute shite, and giving them any credibility, giving any indication that the WMF/Wikipedia things it's a good thing that people use them, if they just add Wikipedia as a "preferred" source, is sending out the completely wrong message. Google AI is something which should be opposed where possible and ignored otherwise, not in any way enouraged. Fram (talk) 08:41, 10 September 2026 (UTC)

Whether it's a good thing or not people do use them. Google itself is shit. We've been integrated with Google for years, I do not think that is an endorsement but a reflection of the realities. Are we going to block ourselves from Google to maintain the moral high ground? PARAKANYAA (talk) 13:31, 10 September 2026 (UTC)

We don't present all readers coming from Google with a "hey, you can use "Wikipedia on Top" to always get Wikipedia as your first Google result either", we don't comment on what search engine people use and don't in any way indicate that we think this was a good choice or how they can make it even better ("hey, you can put DuckDuckgo as your preferred search engine!"). My reply doesn't "block ourselves from Google", plrase don't use such strawmen. My reply avoids us encouraging readers putting a flag on a turd. Fram (talk) 13:44, 10 September 2026 (UTC)

Great care would have to be taken to present this as a dry technical option than something being pushed. CMD (talk) 13:52, 10 September 2026 (UTC)

Disagree, I see no problem with encouraging people to do this. I do think we should be careful not be annoying in the way the option is presented (and to avoid banner blindness). Levivich (talk) 13:55, 10 September 2026 (UTC)

Sounds like a good idea to me, at least to run the experiment and gather the data. I don't agree with Fram that we should oppose Google, or any other search engine. Wikipedia doesn't need to take a position regarding what other tools readers use. Personally, I don't think we should care about clickthroughs, but we should integrate with other software that readers use when it makes sense to do so. This is one of those opportunities, why not make it easier for people who want to find Wikipedia pages to find Wikipedia pages. Levivich (talk) 13:53, 10 September 2026 (UTC)

I don't oppose Google, I oppose the Google AI results which appear at the top of their searches and are among the least reliable LLMs you can find. Fram (talk) 08:02, 11 September 2026 (UTC)

Yeah, free Gemini sucks. I still think we should try out Google's preferred sourcing setting even though free Gemini sucks. Maybe especially because free Gemini sucks: help those poor googlers out. Levivich (talk) 23:46, 11 September 2026 (UTC)

Sounds like a great idea! (I've seen this previously on Wikipedia:Discord). However, what if people accidentally close the toast, but then decide that they want to add Wikipedia as one of their preferred sources? Perhaps that could be added as a menu item below the page title, or the navigation bar at the left side? Also, how can us logged in editors try this feature out first-hand? And just to clarify, has the "Wikipedia articles showing up on the Google app's feed" started rolling out? Hason-LEK-🇸🇬 ● 💬 ● 🧱 ● 🧪 02:47, 11 September 2026 (UTC)

Sounds great! I had already set Wikipedia as one of my preferred sources last month. FaviFake (talk) 08:03, 11 September 2026 (UTC)

I do share many of Fram's and CMD's concerns here, and I think we will need to consider carefully whether this is something we want to do long-term. The experiment will supply data that can be useful in those discussions, and the ethical questions are important but not that urgent, so that I do think the experiment could/should go ahead. ;; Maddy from Celeste (WAVEDASH) 19:44, 11 September 2026 (UTC)

Big thank-you to all who have weighed in so far! I added a note about this proposal to the monthly installment of Tech News (going out next week) to make sure it's visible to any others who might have an opinion. Surely we can get at least as many participants in this discussion as the spirited debate above about removing Hersheys logos from WikiLove templates... Happy Friday, and thanks again for your thoughts here! Talk more next week. Maryana Pinchuk (WMF) (talk) 21:46, 11 September 2026 (UTC)

You should let Hersheys sponsor this feature! It will be the discussion of the century! FaviFake (talk) 21:59, 11 September 2026 (UTC)

I don't think this is a good idea. 'Preferred sources' are for the AI overview, and I don't think anybody has ever actually clicked on the sources to read on further from the AI overview (speaking from experience watching others use it, I haven't used it enough myself to tell). This probably wouldn't benefit Wikipedia that much, and on the other hand people who would see this and actually care about reading the sources might think that Wikipedia is AI slop or something. So, I think the cons outweigh the pros. Also this might be a privacy concern? Not sure though. -I sometimes eat bananas, and you can talk to me here: (talk) I sometimes eat bananas, and you can talk to me here: (talk) 22:12, 11 September 2026 (UTC)

Google is shaping how people access the Internet, and there are a lot of different uses of the AI overview. In my personal circles, people at least sometimes click on the links to double-check what the AI wrote (and, frequently, learn that the AI was being confidently incorrect). It's actually surprising to me Wikipedia is not already preferred by Google, since it's being scraped by LLMs multiple times a second. This feature seems to me like a strong benefit for general audiences with little downside. It's worth testing out at least. NotBartEhrman (talk) 00:50, 12 September 2026 (UTC)

Support this. I disagree that we shouldn't give recognition to Google AI, because even though we might not like it, the truth is that lots of people use it. And yes, I think some people actually do click the sources in Google AI, albeit not everybody. Axolitl (talk | contribs) 02:57, 12 September 2026 (UTC)

I support this. I do not think the counter arguments hold water to anyone that is pragmatic Czarking0 (talk) 05:55, 18 September 2026 (UTC)

Pinning edit requests

Should we pin unanswered edit requests so they don't go into the archives? The reasoning is that there are quite a few unanswered edit requests that just go straight into the archive with no one responding. (example 1, example 2 (nine of them!), sort of example 3) AlphaBetaDeltaLambda(αβδλ)^talk 18:05, 13 September 2026 (UTC)

This seems like a good idea. I wonder if this could be built into the templates directly so that the pin is removed when |answered=yes is entered. —Myceteae‍🍄‍🟫 (talk) 18:13, 13 September 2026 (UTC)

Not entirely sure that's possible. {{DNAU}} works by placing an HTML comment on to the section so bots know not to archive it. It might have to be a bot. AlphaBetaDeltaLambda(αβδλ)^talk 18:16, 13 September 2026 (UTC)

That, or updating the code of archival bots. Axolitl (talk | contribs) 05:18, 14 September 2026 (UTC)

• Meh… I think that if an edit request has been sitting unanswered for a long time, it is highly unlikely that the request is ever going to be answered. There comes a point when we should automatically mark it as “not done” and close the request.

We can discuss how long we should wait before such an automatic close, but it would be ridiculous to pin a request that has gone unanswered for (say) a year. Blueboar (talk) 13:40, 17 September 2026 (UTC)

I could support some sort of automatic process for marking these not done. I think the documentation should make it clear that the request has expired, as opposed to an editor making a determination to not implement the change. I don't monitor edit request backlogs and make edit requests infrequently, so I'd want to hear from more folks who are active in this area about a potential change. You're right that it doesn't make sense to leave these open for many months but it would be better to have a default standard rather than depend on the auto-archive settings on each page to determine when to disappear them from view while still leaving the request officially open on the template. —Myceteae‍🍄‍🟫 (talk) 14:18, 17 September 2026 (UTC)

Yes… Automatically closing (and archiving) as “Expired” is a good approach. Doesn’t come across as “biting” the editor making the request, but allows us to clear the backlog at both the template and the talk page. Blueboar (talk) 14:32, 17 September 2026 (UTC)

Yes, not bitey, and just more accurate to reflect the actual outcome (or non-outcome). It could be documented as |answered=expired or with a separate parameter and the displayed text could make it clear what the status is. I wouldn't automatically archive them but would let the workings of the talks page handle that. Expired edit requests don't need to be archived any faster than other posts on the same talk page that get no response. —Myceteae‍🍄‍🟫 (talk) 14:50, 17 September 2026 (UTC)

Sounds right to me. I was assuming the issue was with archiving requests that were still open… not when archiving takes place. so my suggestion was - IF a request is still open when it would normally be archived, automatically close it as “expired” and then archive as normal. Not faster or slower. I think we are on the same page. Blueboar (talk) 15:43, 17 September 2026 (UTC)

I'm still sympathetic to the nom's initial proposal. I'm not sure where I stand. We tell people that it can take anywhere from a day to several months for an edit request to be acted upon although most of the longer timelines are for COI requests. If non-COI requests are regularly taking >3 months to get any response, that's not great. Of course the requesting editor is supposed to take some responsibility to appropriately move the request forward and if they've moved on in that time then there's little sense in leaving the open request in the backlog. —Myceteae‍🍄‍🟫 (talk) 16:13, 17 September 2026 (UTC)

(pinging editors who may be active in answering edit requests: UmbyUmbreon, Day Creature, AlphaBetaGamma, Deacon Vorbis, I am bad at usernames, LizardJr8)

Yeah, COI edit requests are really backed up (there are around 900 COI requests, compared to 120 extended protected and 40 semi protected), mostly because very few editors wants to read a wall of text. I don't think a requesting editor would be encouraged though if their request just "expired" though; it seems like pushing our (the reviewers) problem to them. Although I agree that requests can't be left forever. Perhaps auto-expire requests from inactive editors but keep active ones? AlphaBetaDeltaLambda(αβδλ)^talk 16:50, 17 September 2026 (UTC)

I can only speak for protected requests as I don't really touch the COI ones much. Generally, if a request has been sitting for weeks or months it's unlikely that it will ever get answered. Usually this is because of issues like the request being a wall of text, too complex, or requiring specialized knowledge or hard-to-access sources to evaluate. Putting some kind of expiry mechanism in place seems like it could be helpful, both in clearing out the backlog and in letting any of the requesting editors who are still around know that there was some kind of issue with their request and they need to try again. Not sure of what the exact timeline should be or how to phrase the "request has expired" message to be most informative to the requesting editor. Day Creature (talk) 17:23, 17 September 2026 (UTC)

Rather than keying it on expiring a time window, my view on the old ones is that there is no consensus to adopt them.

The COI ones are often so long, convoluted and nitpicky that they are terrible to clear. (My fondest wish is that there was something that indicated the level of change, because an updated sentence or two is easier to deal with than "the complete rewrite in my sandbox".) LizardJr8 (talk) 18:11, 17 September 2026 (UTC)

Wouldn't a default expiry time window be one way to document this? For any long-unanswered post, a reasonable interpretation is that there is no consensus though determine what to do with it requires some judgment. It's one thing for informal, free-texted proposals to go stale and fade away but for formal processes that generate a backlog it seems desirable to:

1. Not prematurely archive live requests that potentially could our should be addressed;
2. Not mark them open indefinitely and maintain a large backlog of very old requests that are unlikely to be responded to; and
3. Have more clear documentation of the status of such requests.

We have default actions to be taken after a standard timeline for other processes like WP:PROD, RM (WP:RMNOMIN), and XFD (WP:SOFTDELETE). There are a number of important differences here, of course. —Myceteae‍🍄‍🟫 (talk) 18:59, 17 September 2026 (UTC)

Yes, I articulated myself poorly... I was thinking that the closure could be time-based, but the stated reason can be a lack of consensus based on the time elapsed without action rather than just saying "it's old".

I agree with your points. LizardJr8 (talk) 19:20, 17 September 2026 (UTC)

The wording would need to be workshopped. We could adapt it from the current guidance at Wikipedia:Edit requests § Unanswered requests as well as § Response time further up the page. And we could update the wording on the information page to reflect any process updates related to expiration. —Myceteae‍🍄‍🟫 (talk) 20:03, 17 September 2026 (UTC)

Eh, my general gut feeling here is that the archiving bot(s) simply shouldn't archive a section that has an unanswered edit request template in it. Sometimes this happens for very stale requests that are unlikely to really get addressed. If you want to automatically close those (or bundle that feature in with the archiving bot), that's kind of a separate question that I don't feel strongly about. Occasionally this can be more of a problem on very active pages that have a high volume of new sections and a low archiving delay. That's a problem when it happens, but I only see it very rarely. –Deacon Vorbis (carbon • videos) 16:57, 17 September 2026 (UTC)

Pinging Rusalkii who has given us User:Rusalkii/Responding to COI edit requests. —ClaudineChionh (she/her · talk · email) 00:36, 18 September 2026 (UTC)

Optimized user signature

This is a minor thing, but in my opinion makes everything less convoluted and easier to see in source editor (which is forced in discussion pages). The default username signature is [[User:Example|Example]], which displays as Example. However, if you remove everything after pipe, it uses the first pipe and omits everything before a colon. So if you make your signature [[User:Example|]], it will also be displayed as Example. It won't be perfectly accurate all the time, like if a user has brackets or commas in the username, but these rare cases can be easily indentified. So something like "if brackets or comma, duplicate name after colon; else nothing after pipe". It's not the highest priority to save a few kilobytes in discussions, but still makes it easier to see in source editor. Dabmasterars [RU/COM] (talk/contribs) 16:27, 18 September 2026 (UTC)

(This is mentioned at Wikipedia:Pipe trick.) LightNightLights (talk • contribs) 00:16, 19 September 2026 (UTC)

It has advantages but it doesn't save any bytes, because User:Example| is expanded to User:Example|Example when the page is saved, so the full version is stored in the wikitext. Certes (talk) 09:45, 20 September 2026 (UTC)

I just opened this page in source where I used the trick, and yeah, you're right. Well, in that case, I want this not to happen, so that instead of adding it after a save, it just renders it without bloat, just like suffixes after square brackets work. Dabmasterars [RU/COM] (talk/contribs) 10:07, 20 September 2026 (UTC)

That would require a software change, which is something that would need to be requested at Phabricator (see WP:Phabricator if you're not familiar). Thryduulf (talk) 12:43, 20 September 2026 (UTC)

Right, I was half-expecting to be posting in the wrong place, thank you for directing me. Dabmasterars [RU/COM] (talk/contribs) 14:33, 20 September 2026 (UTC)

If deployed from Beta, disable Image Carousel for logged-out users by default

• Prior discussion: Wikipedia:Village pump (policy)/RfC on the new Image Carousel feature

Now that the RFC discussion is closed as deeming the Image Carousel feature noncompliant with WP:IUP and WP:NFCC, the right thing to do before deployment from Beta is disabling/opting-out the feature from logged-out users, including TA users. Whether to hide/suppress the feature from such users can be also discussed here. Indeed, I'm doubtful that disabling the feature means hiding or suppressing it also. If I see the consensus favoring this proposal, then I would create the Phabricator task requesting such. George Ho (talk) 01:11, 19 September 2026 (UTC)

Should it be made impossible to create an article when there is a draft with that exact name?

I don't know how it would work on a technical level, but wayyyyy too many people ignore the pop-ups in big letters and a striking red background which say there's already a draft. ~2026-41404-95 (talk) 03:43, 20 September 2026 (UTC)

If a vandal or incompetent editor creates a hopelessly bad draft named for a notable topic, I don't see why that should prevent someone with a clue from creating a similarly titled article properly. Certes (talk) 09:36, 20 September 2026 (UTC)

I don't see why the vandalised/low-quality content couldn't just be replaced with good content followed by the draft being accepted. ~2026-41404-95 (talk) 09:42, 20 September 2026 (UTC)

It could be that the draft is about an entirely different topic. Thryduulf (talk) 12:44, 20 September 2026 (UTC)

I'm struggling to think of an example. If an article about an entirely different topic could plausibly have an identical name in draftspace, you could argue both should have a disambiguator ~2026-41404-95 (talk) 12:55, 20 September 2026 (UTC)

That presumes the draft is on a subject which has a valid claim to notability. For example we shouldn't require disambiguation for a biography for a clearly notable individual just because someone has created a draft on another individual who shares the same name, but may not actually meet notability criteria. Things that exist in draft space should have no bearing on article content decisions - which include article titles. AndyTheGrump (talk) 13:23, 20 September 2026 (UTC)

Even if the draft is about a different notable topic, the title of the existing article should not be disambiguated unless it is ambiguous with another extant article. If the draft is accepted then the article that was created first should (and normally will) be moved to a disambiguated title at that time (if it is not the primary topic). Thryduulf (talk) 13:39, 20 September 2026 (UTC)

An editor capable of creating an article in article space would also be capable of moving a draft in draftspace to a disambiguated title. I would say, 1) move the "bad" draft on the different subject, then 2) create the new draft for the new subject at that title at the draftspace title vacated by that move, and then 3) move the new draft to mainspace. There are two extra steps but the overlap problem is solved and the other draft is now pre-disambiguated. BD2412 T 19:27, 22 September 2026 (UTC)

One question is who should get "page creator" credit, which for example yields notifications when a link to a page is made (and is used in some internal challenges). A user writing a mainspace ready page should be preferred over the creator of a substub draft. —Kusma (talk) 14:09, 20 September 2026 (UTC)

Regarding notifications when a page is linked to, you may be interested in phab:T66090 which is a request to allow users to specify an arbitrary list of pages they get such notifications for, not just those they created. This has been open since 2014 though so it is not likely to arrive any time soon. Thryduulf (talk) 14:28, 20 September 2026 (UTC)

No, it should not be made impossible. Would you also make it impossible to move userspace drafts over to mainspace if there is an existing draft? —Kusma (talk) 14:06, 20 September 2026 (UTC)

I'm inclined to say no, let competing editors do their thing. — VPP (t/c) 18:38, 22 September 2026 (UTC)

RFC on adding birth_name parameter to sports bio template

Input is needed at Template talk:Infobox gridiron football biography#RFC on adding Birth_name parameter. This issue failed to gain attention when it was raised last year by another editor, so I started an RFC to get more input from the community so that a WP:CONSENSUS opinion could be reached. Thanks to anyone in advance who is just willing to show up and make a comment. All opinions welcome.4meter4 (talk) 17:39, 21 September 2026 (UTC)

Your feedback wanted: Remove (Language link added:) and similar edits from Special:RecentChanges

Hello, I’m Danny from Wikimedia Deutschland’s Wikidata For Wikimedia Projects team.

We have been investigating and examining how Wikidata’s use in the Special:RecentChanges and Special:Watchlist pages can be made to be more useful, or less ‘noisy’ (fewer edits that have little value to the reader).

We are currently exploring to make a change in the volume and the type of Wikidata edits that appear in these pages, and we need your feedback to help us ensure this is a positive change for the Wikis and won’t disrupt your workflows, or lessen your experience.

Language links are also known as sitelinks or interlanguage links and describe when a Wikidata item is connected to a Wiki, adding it to the language dropdown tool for navigation between other connected Wikis.

A variety of "Language Link" edits as seen in the Recent Changes feed

The request for feedback is whether you support the removal of these language link edits from the Recent Changes and Watchlist feed, or would you oppose such a change and if so, why?Please share you thoughts on the dedicated metawiki Talk page.

Further information about this change: m:Wikidata_For_Wikimedia_Projects/Hide_Language_Link_edits

Thank you for any participation or feedback you give. -- Danny Benjafield (WMDE) (talk) 15:34, 22 September 2026 (UTC)