Wikipedia:Village pump (miscellaneous)

The new Google search has been announced a few days ago (their blog, but also reported by virtually every major source). This would basically mean that already this year our audience would shrink from being part human, part AI to being fully AI, no humans will read Wikipedia. This probably should have some consequences for what we are doing (you know, tastes of AI and humans are still different, though they seem to be converging very quickly). I have not seen this being discussed in the Wikimedia universe, and I would appreciate some links if someone is aware of such discussions. In the unlikely case it has not been discussed, we probably should start a discussion here. Thanks. Ymblanter (talk) 11:26, 23 May 2026 (UTC)

The Google to Wikipedia reader is not the only type of reader, and they're really not the kind of reader I write for. Thebiguglyalien (talk) 16:06, 23 May 2026 (UTC)

My reading of the blog: "blah blah agentic coding blah blah intelligent search blah blah AI Mode blah blah Gemini 3.5" etc. Really there's nothing new announced that doesn't already exist in Gemini itself. SuperPianoMan9167 (talk) 16:58, 23 May 2026 (UTC)

tastes of AI and humans are still different, though they seem to be converging very quickly That's because you're projecting the tastes of humans onto nonhuman things. SuperPianoMan9167 (talk) 17:01, 23 May 2026 (UTC)

You may be interested in "Emotion concepts and their function in a large language model". Sean.hoyland (talk) 17:25, 24 May 2026 (UTC)

Meh. Still seems to just be anthropomorphism: Note that none of this tells us whether language models actually feel anything or have subjective experiences. SuperPianoMan9167 (talk) 17:29, 24 May 2026 (UTC)

Also, Anthropic has a strong incentive to claim that their models have become sentient, because then they can make money off of marketing themselves as "promoting AI welfare". (Yes, I am an AI skeptic.) SuperPianoMan9167 (talk) 17:32, 24 May 2026 (UTC)

To understand these models’ behavior, anthropomorphic reasoning is essential. is the part I thought might interest you, anthropomorphism as a tool for alignment. Aside from just trying to understand what is happening inside these networks with their interesting circuit structures, it seems to be about finding ways to align the values and preferences that influence their behavior with ours. I'm a skeptic by nature and training, but I think it's okay to use familiar words as stand-ins for emergent behavior that we don't understand produced by a bunch of vectors. Sean.hoyland (talk) 18:28, 24 May 2026 (UTC)

I mean, of course it isn't sentient. It's just predictive. Hypothetically, if you understood all of the weights, you could emulate chatGPT with pen and paper. Nobody would argue that's sentient lol. Gaismagorm (talk) 13:04, 2 July 2026 (UTC)

My understanding they are going to get rid of the search feed, so that people would not be able to get to the websites in one click. This pretty much means they will not be getting to us. There are ways around this, for example (just throwing up smth) supporting DuckDuckGo which is currently does not provide the search of a quality comparable to Google. I am sure there are other solutions, but if we do nothing we just lose 99% of our direct audience. Ymblanter (talk) 17:32, 23 May 2026 (UTC)

It seems timely to explore the consequences if Alphabet abandon the longstanding arrangement that Google indexes and presents Search links to websites.

While incoming traffic to WP and other sites is the first victim, there are other consequences. Take the longstanding understanding that new unpatrolled pages are unindexed; avoiding hoax/nonsense from being surfaced was in everyone's interest when search accuracy was the thing, but in this new sloppy world, is that exclusion being respected by LLM bots, including those using the Enterprise API?

And for editors here, Findsources for Notability checking is primarily linking Google Search. Is that sustainable if it returns Gemini slop? Would appending udm=14 maintain the prior norm? Or should it be parameterized to the user's preferred search engine? Or seek a Wikipedia Library type arrangement with Kagi, for example? AllyD (talk) 11:13, 24 May 2026 (UTC)

Thanks. This is indeed one thing, another thing is that whether a page looks nice to a human (and whether it is the case for desktop/mobile, different resolutions etc) becomes largely irrelevant, only the content. May be not even interlinking the pages. Ymblanter (talk) 16:04, 24 May 2026 (UTC)

And, as another consequence, I understand that it looks like red tape for some users, but Commons becomes more important under these conditions than Wikipedia (in any major language). Ymblanter (talk) 16:08, 24 May 2026 (UTC)

I disagree, if we did this we would just be exacerbating the problem. I don't think giving up on our human readers is the answer. ~2026-35582-22 (talk) 20:56, 17 June 2026 (UTC)

It says You’ll continue to get a range of results from Search, just like you do today. That does not sound like "they are going to get rid of the search feed, so that people would not be able to get to the websites in one click" to me. WhatamIdoing (talk) 05:59, 25 May 2026 (UTC)

Google for years has been moving away from delivering a list of blue links in response to search queries. But the refreshed search engine, which runs on the company’s new Gemini 3.5 Flash model, represents what may be its biggest shift yet toward AI and away from traditional search? Just a random article from my feed. Ymblanter (talk) 06:48, 25 May 2026 (UTC)

Probably "will not be able" is too strong, if they are determined they will. But most will be not even interested at clicking at the links. Ymblanter (talk) 06:50, 25 May 2026 (UTC)

I think this is the main threat: People will not be interested in clicking on the links, because they'll get their question answered without needing to click. WhatamIdoing (talk) 18:52, 26 May 2026 (UTC)

I thought this is pretty much what I said, but may be I was not clear enough.--Ymblanter (talk) 06:43, 27 May 2026 (UTC)

Yes, exactly that. "When people search with AI, they're less likely to click through". Google's AI Overviews are already causing Wikipedia to lose traffic. If and when Google makes "AI mode" the default search mode (or makes searching on Google similar to chatting with an AI chatbot), then that'll pretty much spell the end of this project. Some1 (talk) 00:50, 27 May 2026 (UTC)

My reading is this is what they are going to do this Summer, and this is not yet the end of the project because the LLMs need to get the information elsewhere, and Wikipedia is still the best source of structured information which is still being added on a regular basis, but, as I said in the opening statement of this thread, we will have to adapt to the fact that we will not get any human readers, only LLM readers.--Ymblanter (talk) 06:47, 27 May 2026 (UTC)

They're not getting rid of the search feed. They're introducing "agents" that will "autonomously crawl the web on your behalf". SuperPianoMan9167 (talk) 14:55, 25 May 2026 (UTC)

AKA Google's own version of OpenClaw. SuperPianoMan9167 (talk) 14:58, 25 May 2026 (UTC)

My main worry is how AI manipulates information on controversial topics. I have tested this with queries like "Gaza genocide" at various times in the last 2 years. I prefer just regular web search and access to the open web. But in 50 years, the new generation might not know what the open web is.... 🐈Cinaroot   03:44, 26 May 2026 (UTC)

According to Google's own AI, "Users, tech critics, and researchers have documented a measurable decline in Google Search quality ... from a mix of aggressive monetization ... and the disruptive introduction of AI features." Certes (talk) 15:29, 26 May 2026 (UTC)

I think the complaints about Google Search are correct, but I think it's more complicated than that. I've used DuckDuckGo as my main web search engine for a few years. I find that the quality is adequate for most everyday, low-stakes purposes (e.g., finding a business's website or looking up a word/place/product that's mentioned in something I'm reading). But I find Google to be better if I'm looking for more complex things (e.g., a specific news article whose title I have, finding something relevant when I ask for information from slightly wrong keywords). What's annoying with all of them is that when I search for exact phrases (e.g., a quotation from a source), they may not find the thing that I want, and they always throw in things that I don't want.

Earlier this month, I switched to having Wikipedia be my default search tool, and I've been surprised at how well that's going. WhatamIdoing (talk) 19:13, 26 May 2026 (UTC)

That's an interesting idea, though obviously not great for finding reliable sources. I use DuckDuckGo (Bing with a false beard) but it's also going down the AI route, presenting "Learn about..." summaries as if they were Wikipedia leads. Certes (talk) 09:17, 27 May 2026 (UTC)

Humans will continue reading Wikipedia, in the same way they will continue reading many other websites (the sites that will be really damaged by this will be minor ones that almost nobody knows, and it's a pity anyway).

Wikipedia is the largest reference work ever written, and also the most read one (yes, this will continue to be so, since LLMs and search engines are not written works). Even if, eventually, most people don't want to read more than three lines of AI-generated content, the ones who choose to retain their brains despite having AI available, would want to continue reading, and Wikipedia is a very interesting work to be (of course, partially) read.

AI summaries are very useful to get quick facts, and they are no problem, since WMF's money doesn't come from clicks. When you need detailed information or a permanent written work whose content doesn't change every time you perform the search, you click on the link (that still exists, even in Google's AI mode) and come here.

Two Wikipedia essays that I think can help you to highlight why Wikipedia is not just a tool to search for information, as search engines or AI-powered search are, but it is information in itself (disclaimer: they were both written mostly by me):

• Past and future of Wikipedia
• The hidden encyclopedia that resides in the article histories

MGeog2022 (talk) 12:27, 2 June 2026 (UTC)

I thought I should link this meta-wiki page here, it is very relevant.

https://meta.wikimedia.org/wiki/Wikimedia_Foundation_Annual_Plan/2026-2027 ~2026-35582-22 (talk) 20:52, 17 June 2026 (UTC)re

Just to note that there is currently a discussion on the wikimedia-l mailing list (which I guess most people thought disappeared in the end of the 2000s). It is unfortunately spread over several threads, but what I want to note here is that LuisVilla, addressing the situation, calls it "Google Zero".--Ymblanter (talk) 06:46, 8 July 2026 (UTC)

It always bothered me that among millions of articles, there might be stuff that we missed with our usual tooling (NPP for example) and they might lurking somewhere while this is not really a hard problem to solve. Also, NLP is a hobby of mine (1 2) so I built a small tool that can take a wiki, take all the articles and orders them based on how similar they are to members of a given category but dissimilar to members of another category (you can generalize this even further but let's stick to this for now).

So for example Special:PermaLink/1361555362 is list of articles that are the most similar to articles tagged with {{Promotional}} but furthest from Featured/Good articles. A lot of them clearly need some clean up but nothing egregious thankfully. I would be grateful if people take a look at those. I think if I can get my hands on a corpus of deleted articles for G11 (I'm not admin, so I don't have access to those). I could find more blatant promotional articles too.

I can try it with other categories as well. Like the {{Tone}} maybe. If we were a startup and wanted to use marketing, we could claim "We use AI to find issues in Wikipedia!" since strictly speaking NLP is part of AI but it's galaxies apart from what people think of AI these days. It's not LLM nor generative AI and as such resource consumption of it (electricity, hardware, etc.) is extremely low.

What do you think? Should we do more of these? Have other people tried this before?

Technical details so feel free to skip this paragraph if you're not interested (for people who do NLP, feel free to give feedback to make it better): Using CBOW, I produce a word2vec with vocab size of 3000 and embedding layer of 100. Then go through each word in every article and using the word2vec mapping produce that many vectors, the average the vectors to produce a vector representing the article. Then we go through members of several categories. Promotional one and then average them to produce the "center" of promotional content, and similarly for featured and good articles to have a couple of "centers". Then compute the cosine similarity of each article with tone center minus cosine similarity of featured centers (I take the lowest distance) and order them from lowest to highest and take first N results. I can explain other methods I tried and the results were not good and why but that'd be rambling :D Ladsgroup^overleg 18:40, 28 June 2026 (UTC)

For example: Silvia Lenaerts: "She is a strong advocate of cooperation between universities, government, and industry in finding solutions to societal problems." this seems very non-encyclopedic :D You can find a lot more there. But obviously, similar to any statistical tooling, it will have false positives.

Oh and also, I can run this for any language basically. If you're active in another wiki and want to get this for yours too. I can try to do it. Just send me an email. I already did this for Persian (and the result were quite good there, much better than enwiki). Ladsgroup^overleg 18:46, 28 June 2026 (UTC)

That could definitely be interesting! Feel free to suggest this at Wikipedia talk:WikiProject AI Tools, this certainly seems like a good use case for AI help. Chaotic Enby (in solidarity · talk · contribs) 19:06, 28 June 2026 (UTC)

Thanks! I mentioned it there. Ladsgroup^overleg 20:43, 28 June 2026 (UTC)

This is very neat! An interesting type of article to find would be articles similar to {{COI}} and {{Undisclosed paid}}, but not tagged with either of those tags, to seek out undisclosed paid edits that have evaded notice. For a public corpus similar to articles deleted for blatant promotion, you could use AfC submissions declined as an advertisement and maybe AfC submissions rejected as non-notable.

I think there's a lot of potential benefit in building and sharing good tools for prioritizing articles that need improvement within a category or a WikiProject (some notes on this topic). For example, on Wikipedia:WikiProject Computer security I've linked to a an auto-generated list of articles in scope that need cleanup made by User:CleanupWorklistBot, which is very helpful, but there are almost 2000 articles on that list! The only quality metrics that this list offers are the number of tags and manual assessments, which are of limited value. My current approach is to take that list and process it with the prototype at Projo, which incorporates pageview counts and LiftWing article quality scores, to get this ranked and filtered list. It'd be interesting to add automated scores for various kinds of cleanup issues, like COI and buzzwords, not just a single quality score. The Microtask Generator takes another approach with displaying quality scores with facets/sub-scores for different kinds of article structure issues.

Also appreciate you sharing your method, since I've done a bit of work with vector embeddings and understand some of what you mean. :) (For anyone else interested: I found the Gensim intro exercise super helpful and fun, and it only requires very basic Python skill.) Dreamyshade (talk) 01:25, 29 June 2026 (UTC)

Also, processing your list through Projo produces interesting results. For example, with that method of ranking, I looked into the top article (Srini Raju) and found that it completely omits mention of a scandal that person was associated with. Dreamyshade (talk) 02:00, 29 June 2026 (UTC)

Category:Articles containing suspected AI-generated texts is another obvious category to target, for the selfish rationale that it would be nice for someone other than me to be finding this stuff. Though there is a slight problem in that it lumps articles from 2023-2026 together, and AI output sounds very different across versions and years, and measurably different from each other. (I need to re-run this as the corpus is much larger now but User:Gnomingstuff/AI experiment/New AI versus old AI gives you an idea)

(but I do have a private corpus of AI-generated articles and article excerpts, sorted by year, origin, type of article if it's a human-written one, etc.; I can send it to you if you want) Gnomingstuff (talk) 14:41, 29 June 2026 (UTC)

Thanks! Looking at the rejected AfCs sounds like a good idea but it's bit of work since it's not in the vector dataset I built. It's not hard to make it to work though. I just need to it on the weekend. The Category:Articles containing suspected AI-generated texts is a good one and the articles should be in the dataset already too. I run it now and report back once done. Ladsgroup^overleg 15:28, 29 June 2026 (UTC)

@Gnomingstuff Here are the result: Special:PermaLink/1361762830 (euclidean distance) and Special:PermaLink/1361762710 (cosine delta). Let me know which one ends up being better! Ladsgroup^overleg 23:10, 29 June 2026 (UTC)

Thanks, will take a look once I get a chance! Gnomingstuff (talk) 23:21, 29 June 2026 (UTC)

The Euclidean distance list seems more effective to me. The cosine delta option seems to make a list of articles that are under-developed in general, rather than having LLM issues in particular. Found a few articles to clean up: Al Akhawayn University (not tagged for LLM content), American Councils for International Education (not tagged for LLM content), and and Ministry of Culture (Russia) (already tagged for LLM content). Also interesting to compare the results in these lists to the scores at https://wikipedia.gptzero.me/.

Just read this paper released last month about detecting LLM text in Wikipedia articles (with a WMF researcher involved). For some reason I thought that the Wikimedia Enterprise API already had a credibility signal about likely LLM content, but I'm not seeing anything like that in the LiftWing docs. Dreamyshade (talk) 00:52, 30 June 2026 (UTC)

whoa since when did this gptzero thing exist Gnomingstuff (talk) 02:36, 30 June 2026 (UTC)

also I'm going through them here but one at a time: User:Gnomingstuff/AI/Article similarity audit; some of the maybes are probably yes but I am being conservative with handing out yeses, basically it's "would I tag this if I saw it in the wild". seems to be really indexing on schools/organizations for some reason Gnomingstuff (talk) 02:38, 30 June 2026 (UTC)

@Ladsgroup Finished going over these (at the above "Article similarity audit" link). Not really all that much of a difference between the two lists, both are ~15% stuff I would tag as AI and ~17% stuff that might be AI. Will say that my gut instinct was almost always correct. Gnomingstuff (talk) 20:07, 2 July 2026 (UTC)

@Gnomingstuff Thanks for going through the report. Do you find it useful or time-saving so I would produce it from time to time or you'd think you have other means to find such articles. I can for example add an extra filter to just discard any article that the text is mostly unchanged since 2020 which should take care of most false positives. Ladsgroup^overleg 15:07, 3 July 2026 (UTC)

The filter would help a lot! Gnomingstuff (talk) 15:25, 3 July 2026 (UTC)

@Ladsgroup Filtering out articles with text mostly unchanged since 2020 sounds helpful! Often LLM content is added to existing articles as a new section, so I wonder if there's a good opportunity to apply chunking strategies to find and highlight that kind of material.

To me, any kind of automated detection of potential LLM content for editor review is super helpful to see in real life outside of academic papers, and 15-17% success rate is a good starting point. Other contributors could help iterate on this too. Do you have your code open source somewhere? I'd be curious to try running it scoped to articles in a category or WikiProject. I'm also helping a bit with the Wiki AI pre-conference day at Wikimania, and I'd like to encourage people to check out the code and try it out - it'd be a nice way to encourage recognition of NLP/ML strategies as useful, efficient, fun tools even in an era of LLMs.

Might also be interesting to post on Wikipedia:Bots/Noticeboard and see if somebody would be up for incorporating this into a bot they run? Would be cool for editors to be able to opt into things like:

• Periodic delivery of AI cleanup suggestions on their talk page or user subpage, similar to User:SuggestBot
• Embedded reports on WikiProject pages, like User:Community Tech bot

Dreamyshade (talk) 15:56, 3 July 2026 (UTC)

I'll be also in Wikimania, we can work on it together. Ladsgroup^overleg 19:03, 3 July 2026 (UTC)

The main limitation is the over-indexing on organizations and schools, but I'm not sure what one would do about that. (I guess there are worse potential things to index on.)

I keep meaning to put together a tool for automate finding articles to review rather than having to manually do it across various searches. Even just something like a glorified points system based on this. Gnomingstuff (talk) 19:40, 3 July 2026 (UTC)

@Gnomingstuff That is a really good guide! I used it to try making a little tool that does the search + scan steps. For now it's just a Python script that you can run locally, but I'm going to try to put it on Toolforge. I'm curious what you think! Even just while testing it, I found a bunch of untagged LLM content to remove: Health insurance cooperative, History of tea in India, Body memory, Erovnuli Liga, Georgian Cup. Dreamyshade (talk) 05:03, 5 July 2026 (UTC)

Seems to be promising so far! (Of course anything would need manual review, not just flagging because the tool turned it up, but you know that already.) Gnomingstuff (talk) 18:18, 5 July 2026 (UTC)

@Gnomingstuff Great! Here's a prototype web app version: https://turbo-bassoon-5pgrg497f49rr-8501.app.github.dev/. Still working on the language/instructions to make sure that it gives the right guidance. Up for suggestions. Dreamyshade (talk) 22:49, 5 July 2026 (UTC)

nice! need to remember to make a separate GitHub account so I can push updates if that's ok Gnomingstuff (talk) 05:51, 6 July 2026 (UTC)

Yes, please go ahead and file pull requests! I applied for a Toolforge account so that I can put it up there when ready. The free GitHub Codespaces environment sleeps the app after a couple hours, so that URL will probably 404 sometimes, but anyone can make their own instance of the app in Codespaces. Dreamyshade (talk) 05:58, 6 July 2026 (UTC)

Hey, a finally good use for AI! If this gets added as a tool, I might honestly start using it, despite my extreme anti-AI stance, since I can't see any way this would be abused nor can I see any way that this would involve me outsourcing my thinking to a computer. Good stuff! Hope it goes through! Gaismagorm (talk) 13:00, 2 July 2026 (UTC)

@Gaismagorm Would you be up for trying out this tool for finding potential LLM-generated text: https://wikitomte.toolforge.org/? It's an experiment I made based on the above conversation, and I'm curious if it's helpful for people interested in casually helping with AI cleanup. Dreamyshade (talk) 03:24, 11 July 2026 (UTC)

Perhaps at some point today I'll check out, sound pretty useful! Gaismagorm (talk) 13:08, 11 July 2026 (UTC)

I am not sure where to put this, but here it is.

Several topics related to "user engagement" and the growth team's efforts have appeared in various places lately.

I will keep this short:

If readers spend less time "engaging" with Wikipedia as time goes on, that is a GOOD thing. It means that our information is better written, easier to understand and digest, and the readers are getting what they need here and then moving on with their lives, with new knowledge.

We should not take measures to try to increase the amount of time that readers spend here.

Time spent on the site, per reader, might be a false measure of "success".

Thoughts? David10244 (talk) 05:06, 1 July 2026 (UTC)

I would like to ping some folks from the WMF to hear their thoughts, but maybe they will stumble on this... 😀 David10244 (talk) 05:09, 1 July 2026 (UTC)

It's probably hard to get a clear message of why a reader quickly bounces from a page. Sometimes I am looking for one thing, find it, then leave. However, there are reasons to linger, such as the Wiki rabbit hole (better illustrated at xkcd:214). There may be metrics that better capture different uses. CMD (talk) 06:08, 1 July 2026 (UTC)

You're right that time spent on site could mean many things but I'm hard pressed to see "success" as the best explanation. I know that WMF and the Growth Team look at many different measures and considers various contributors to each of them, though I don't follow this closely. —Myceteae‍🍄‍🟫 (talk) 15:55, 1 July 2026 (UTC)

This is indeed a thorny problem: how do we define success? For most websites, the fundamental goal is to make money, and there's obvious ways to measure that. If you're selling goods, it's how much revenue you brought in through sales. If you're selling advertising, it's how many clicks you generated. For a startup, it's often things that are attractive to the venture capitalists such as Active users.

But we're different. Our product is knowledge, and we give it away. As @David10244 points out, the typical metrics like engagement may not make sense for us. While I get the visceral hate engendered by the LLMs getting fat on our work, is that really a problem? If I write an article which gets ingested by a bot and incorporated into a LLM, and then an end user gets to use that knowledge when they ask a chatbot to do their homework for them, have I not achieved my ultimate goal?

I'm not at all a religious scholar, but I have always been impressed with the eight levels of charity taught by Maimonides. CC-BY-SA-NC seems like the first level (giving begrudgingly). CC-BY is level 5 (giving when you do not know the recipient's identity, but the recipient knows your identity). Having your work spat out by a chatbot without proper attribution gets us to level 7 (giving when neither party knows the other's identity). The problem is, it's really hard to quantify that when it comes time to claim credit on your quarterly OKRs, but if you dig what Maimonides has to say, isn't that pretty close to winning? RoySmith (talk) 15:41, 3 July 2026 (UTC)

@RoySmith Interesting comments. I haven't read Maimonides, but I will. David10244 (talk) 05:52, 5 July 2026 (UTC)

Comment: There is a trend towards people reading less and skimming more, and this is probably an outgrowth of that. Unfortunately, I don't believe we have a new generation of readers that are able to quickly skim and retain information better then previous generations. I would question how much "new knowledge" people are leaving with if they are spending less time engaging with a topic. I've heard several professors complain that students are doing less of the required readings then previous years, and in previous years getting them to do required reading was like pulling teeth. This is a broader societal issue, not one we can really do much about on Wikipedia. GeogSage ^{(⚔Chat?⚔)} 07:01, 5 July 2026 (UTC)

There's nothing new about students being economical with their time by skimping on required reading. See CliffNotes and others in Category:Study guides. Do we not serve that same purpose? We slog through all the secondary sources and summarize the important points so our readers don't have to. And if LLMs then take what we write and further process it into a form that's more convenient, why is this a bad thing? RoySmith (talk) 12:02, 5 July 2026 (UTC)

Sure, but when you have articles like the ones below... It's much more than just "skimping on required reading by using CliffNotes". It's a broader societal issue as GeogSage notes, and AI is only making it worse. There's little Wikipedia can do about it though (unless we completely revamp how Wikipedia articles are written and presented).

• August 26, 2025: Gen Z Is Reading Less. What That Means In The Age Of Ready Answers
• Jan 13, 2026: Gen Z Arriving at College Unable to Read - "It's not even an inability to critically think. It's an inability to read sentences."
• Jan 29, 2026: Professors say Gen Z students can’t read, forcing colleges to lower academic standards - "Professors at Pepperdine and Notre Dame say many college students can no longer read or process assigned texts, increasingly relying on scanning habits and artificial intelligence to get by."
• June 7, 2026: Gen Zers are arriving at college unable to even read a sentence
• June 10, 2026: College Students Are Rapidly Losing the Ability to Read - "There is a measurable, generational collapse in sustained reading and writing." -- Some1 (talk) 15:05, 5 July 2026 (UTC)

Maybe, but maybe all those articles are measuring the wrong thing? Do we want people who can read complete sentences, or do we want people who can find the information they need to function well in society? Back when I was in school and dinosaurs roamed the Earth, the ability to write in cursive was considered an essential skill, and one which I was miserably failing to master. It's a bit of a cliché that adults struggling to master modern technology have to ask their grandkids to help them. Why should we judge what constitutes an essential skill set based on what was essential when we were growing up? Do we really need another generation of people who know how to drive a stick, how to figure out where they are with a sextant, and how to find reference material in a card catalog? RoySmith (talk) 15:36, 5 July 2026 (UTC)

More to the point, most of the world's knowledge is still trapped in old dusty books and periodicals stashed away in library storage rooms, effectively lost to the world. Which is a better way to use that material: wait for wikipedians to grovel over a tiny fraction of that material and hand-craft encyclopedia articles from it, or to launch a moon shot project to get it all scanned and made available on line, indexed by bots, so people can actually use it? RoySmith (talk) 15:50, 5 July 2026 (UTC)

Do we want people who can read complete sentences, or do we want people who can find the information they need to function well in society? Both, and the two skills aren't mutually exclusive. Some1 (talk) 16:00, 5 July 2026 (UTC)

User:Colin Douglas Howell discovered a citation in the article about the early pharaoh Sanakht that badly failed verification (details here). The erroneous citation was on the articles about Sanakht on the German and English Wikipedias by an editor who worked in both languages, and it has since been copied onto the French and Portuguese articles. Howell has corrected the error in the English article, but not elsewhere. For those of us who are not fluent in those languages, is there a better way to notify those wikis of the problem than machine-translating a message about it, posting on the talk pages, and hoping somebody notices? A. Parrot (talk) 19:40, 2 July 2026 (UTC)

There was recently a discussion which inspired the creation of Help:Reporting issues on other wikis. It's quite new and very much a work in progress. Help:Reporting issues on other wikis § Non-English Wikipedias suggests Wikipedia:Local Embassy and some other approaches. Other editors with ideas here should consider contributing to the Help page. —Myceteae‍🍄‍🟫 (talk) 19:51, 2 July 2026 (UTC)

You could also put a note in English on the talk page of the the other language Wikipedia articles, as many users will be able to read English, particularly the person that translated the English version that incorporated the error. Graeme Bartlett (talk) 11:09, 10 July 2026 (UTC)

Dear colleagues, the Wiki Loves Yoruba Heritage photography contest will start on August 1, 2026. We invite everyone to help document and celebrate the rich heritage of the Yoruba people by contributing photographs of our culture, traditions, historical sites, festivals, architecture, arts, and other aspects of Yoruba heritage. To reach more people outside the Wikimedia community, we have proposed a CentralNotice banner to promote the contest and encourage wider participation. Thank you. Agbalagba (talk) 11:06, 4 July 2026 (UTC)

Would someone please delete my own blanked CSS/JS subpages? I needed to create them, because there was no other way to test importing base 64 fonts. The pages meant are:

1. User:Esperfulmo/vector.js
2. User:Esperfulmo/test
3. User:Esperfulmo/common.css/Courier Code regular.css
4. User:Esperfulmo/common.css/Courier Code italic.css
5. User:Esperfulmo/common.css/Courier Code bold.css
6. User:Esperfulmo/common.css/Courier Code bold italic.css

Thanks. --Esperfulmo (talk) 04:15, 7 July 2026 (UTC)

 Done-Gadfium (talk) 04:22, 7 July 2026 (UTC)

Thanks. --Esperfulmo (talk) 04:28, 7 July 2026 (UTC)

Hi everyone. Though I suspect that I'm a bit out of my depth in comparison to some of the more seasoned editors on here, I'm curious what if any mechanism might exist to improve the visibility of RfC topics lacking engagement. Though users will obviously naturally gravitate towards topics that interest them, there seems to at times be a massive disparity in the degree of participation on various questions, seemingly without any obvious rhyme or reason other than two easily observed facts: 1) that many middling topics are forgotten about after a few days, and thereafter buried beneath so much text that users more than likely have to be intentionally seeking it out in order to find it; and 2) that users seem to engage more with topics that already have high levels of engagement (correct me if I'm wrong about this), meaning that topics without significant early engagement rarely end up getting noticed at all. Other than simply leaving the RfC / merge / deletion discussion up for longer in the hopes that someone will eventually find it, what options might exist to notate and highlight the active participation of users in various discussions, outside the talk page itself?

I know that the mobile online version of the site includes a note at the top of every discussion section indicating the most recent timestamp and number of users involved in the discussion. Could something similar be done on the WP:RFC subpages, or at WP:AFD? I'm really just curious what others thoughts are, but will look forward to a fruitful discussion if it arises. Best, and in solidarity, CSGinger14 (talk) 07:51, 9 July 2026 (UTC)

Note that {{WPVG announcements}} does something in the ballpark of what you've proposing, showing the number of participants for video gaming-related AFD discussions. I believe this is maintained by a bot. Mir Novov (contribs | talk) 14:39, 9 July 2026 (UTC)

users seem to engage more with topics that already have high levels of engagement (correct me if I'm wrong about this). I'm not sure about the direction of causality here. A simpler explanation is that discussion topics that are of interest to more editors to begin with are likely to already have a lot of participants at the time that any given editor becomes aware of the discussion, and are likely to continue to attract a lot of new participants, owing to the underlying interest in the discussion topic itself. It's probably a combination of factors. Seeing a large or recurring discussion always piques my interest but sometimes the length and complexity discourages me from participating, especially if it's not a topic I'm already invested in. —Myceteae‍🍄‍🟫 (talk) 17:01, 9 July 2026 (UTC)

Title. I don't know where to put this, so I might as well put it here. Wiki Education is a great idea in theory, but not in practice. Recently, I was editing this article (which I got from Suggested edits): Rezball. It is about a type of Basketball specifically invented by Native Americans. I thought to myself 'interesting, let me do the suggested edit (Fix tone) and be done with it'. I had to remove most of the article and rework much of the content as it was all written like an essay, and not the best worded one at that. The editor who added much of this content was from this Wiki Ed project: Wikipedia:Wiki Ed/University of California Santa Barbara/ENGL 165NJ Native Justice (Spring 2025). Almost every single article assigned to each editor in this project is like this. Practically every editor involved added content that is 1: not neutral, 2: not encyclopedic and 3: written in not the best wording.

I have been noticing a trend lately where articles (especially ones related to Native Americans and other indigenous people groups) have been made exponentially worse by Wiki Ed projects. Most of the time, it seems as if these projects are not supervised or peer reviewed at all, even though peer review is part of most of these project's process. It seems as if people in these projects don't read up well on Wikipedia policies and just give any submitted edits the "A-ok" and move on with their life.

I really don't want to come off as mean or condescending here, but this is a genuine problem I am noticing that, unfortunately, really doesn't have a solution. Any thoughts?

(Side note: when I was starting out, most of my Suggested edits related to revising tone were from articles like this one, I can distinctly remember a really bad one about Native American music) Ilov3gam3z (talk) 17:46, 9 July 2026 (UTC)

I don't think that's entirely fair. Sure, we sometimes have problems with WikiEd projects, but we sometimes have problems with many new editors. It is the nature of learning how to do something that many people will make a hash of their first attempts. If I go back and look at my early contribution history, there's a lot that makes me cringe today. Anyway, @Salo1915 and Brianda (Wiki Ed): who are the two people running that course. RoySmith (talk) 18:08, 9 July 2026 (UTC)

Wrong venue. The proper venue for this is the WP:Education noticeboard, where the broader topic of the effect of student edits is already being discussed in this discussion. Mathglot (talk) 18:11, 9 July 2026 (UTC)

I already have dropped a small comment there, but not many people go to the education noticeboard (most people probably don't even know about it) and I would like thoughts from a larger variety of editors. Ilov3gam3z (talk) 18:16, 9 July 2026 (UTC)

Having this discussion here rather than on the very board dedicated to discussing topics about Wikipedia Education would be an odd choice. And now everyone here knows where the WP:Education noticeboard is. If you have it in both places, it will suffer from thread fragmentation, with people responding in both places, or only one or the other. (edit conflict) Mathglot (talk) 19:02, 9 July 2026 (UTC)

Also, while I would support the discontinuation of Wiki Ed, this is about this problem with Wiki Ed specificly, not anything else. Ilov3gam3z (talk) 19:57, 9 July 2026 (UTC)

I did briefly mention the larger problem in my discussion here, but just in passing. Ilov3gam3z (talk) 19:58, 9 July 2026 (UTC)

Not a subject I know about, but I guess I'll be the one to ask: what were the problems with the student's additions? Yes, plenty of wording/tone issues, awkward headings, and unnecessary external links, but is the article better without the citations and details like e.g. Native Americans were known to have played a sport that was similar to that of basketball, but were not properly introduced to the sport until they were placed in boarding schools by the BIA.[1] The introduction date is marked as having taken place in the late 18th-century early 19th century.[4] (which constitutes about a third of the student's additions right there)? From the headline and tone of this section I expected, well, more than a single example, and for that example to be wholly deleterious rather than just in need of improvement. But again, it's not a subject I know anything about. — Rhododendrites ^talk \\ 19:00, 9 July 2026 (UTC)

I just probably picked a bad example, this was my most recent encounter with Wiki Ed-related problems. Take a look at some stubs/start class articles on niche Native American and in general indigenous people and you will see the problem more acutely. Ilov3gam3z (talk) 19:56, 9 July 2026 (UTC)

Sigh. Okay, so we are fragmenting the conversation. Basically, this has already been asked and answered at the other discussion, albeit to the more general question. However, those responses apply equally here. See, for example, this comment by WhatamIdoing, or my comment there. This doesn't mean that what you are seeing isn't real; it just means you are identifying the wrong bogeyman. I support your efforts and all efforts to improve articles on Native American topics. Mathglot (talk) 21:01, 9 July 2026 (UTC)

IMO this is less a fragmented conversation than it is a perennial one, although I agree with your overall point. The conclusion that ultimately always comes forward is that WikiEd provides a scaffolding that would otherwise make this problem much worse. Criticizing the class and asking to follow up with the instructor and their WikiEd point people is worthwhile; trying to pin the problem on WikiEd itself is not the answer. _signed, Rosguill ^talk 21:05, 9 July 2026 (UTC)

I placed a notice of this discussion at Wikipedia:Education noticeboard. I don't think it's out of scope here, adding it to the very lengthy, months long discussion isn't substantially better, and OP has already commented there on the broader issue. —Myceteae‍🍄‍🟫 (talk) 21:39, 9 July 2026 (UTC)

I think that the OP is confusing the WikiEd program with student edits. We are going to have student edits with WikiEd, or without it. And WikiEd is a strong net positive, given that we will have student edits in either case. And I'll put in my perennial and predictable plug for WP:ASSIGN as useful reading. I don't do any editing on Native American topics, but I suspect that more eyes on the affected pages might be helpful there. --Tryptofish (talk) 22:25, 9 July 2026 (UTC)

I am not confusing anything. There would be less malinformed student edits if there was no Wiki Ed, as Wiki Ed is the main path for these editors to edit in the first place. There will always be these editors; but there could be less. Also, turning this into a 'lesser of two evils' kind of situation is silly and undermines the discussion. This feels like some kind of fallacy to me. Ilov3gam3z (talk) 23:40, 9 July 2026 (UTC)

I think your blame is misplaced. I've run into similar problems, and it usually (not always) comes down to the quality of sources, not the editors, although that can also play a role in controversial topic areas. Viriditas (talk) 01:37, 10 July 2026 (UTC)

As a bit of an aside as I was reading the OP's question, I've noticed classes with the name "Justice" in them seeming to touch on controversial topics (and the capitalized arbcom Controversial Topics) fairly often. Not saying the students are inherently worse in those classes compared to other classes, but picking topics that fall under a CT designation often mean there's a lot more that needs to be checked just due to the combination of student and CT editing. Generally if I see justice in the class name, it's usually a bit more of a red flag to at least check what articles they've been working on as well as checking for advocacy-style editing. In short, your mention of controversial topics definitely illustrates how it can get messy when students get involved even if it's not the students' "fault". KoA (talk) 15:05, 10 July 2026 (UTC)

Yeah. Any Wiki Ed with a loaded word like 'justice' is generally bound to be some form of promotion of a certain view/advocacy. We should really ban Wiki Ed from CTOP, but I digress. Ilov3gam3z (talk) 15:10, 10 July 2026 (UTC)

Since we have a problem with Wikipedia:Systemic bias, then having a class that helps us fix our biases would be a good thing. WhatamIdoing (talk) 19:18, 10 July 2026 (UTC)

These classes dont fix our biases, they just add bias to articles the other way. Ilov3gam3z (talk) 19:30, 10 July 2026 (UTC)

When the problem is missing perspectives, then "adding bias the other way" actually gets the article closer to being neutral. WhatamIdoing (talk) 22:09, 10 July 2026 (UTC)

Two wrongs make a right? Anomie⚔ 22:25, 10 July 2026 (UTC)

No, no, no. Surely you saw the scare quotes, and got the meaning. Mathglot (talk) 22:54, 10 July 2026 (UTC)

??? That's not how things work. Ilov3gam3z (talk) 23:06, 10 July 2026 (UTC)

Let's stipulate for the sake of argument that we would get fewer student editors if Wiki Ed didn't exist. Fine: There would be fewer. But they would be worse. That's not an improvement. WhatamIdoing (talk) 19:20, 10 July 2026 (UTC)

This seems completely wrong. Wiki-Ed as a project has teachers force their students to make changes to Wikipedia. If the project didn't exist these same students wouldn't have any reason to engage with Wikipedia at all. The group of students making Wikipedia edits as extra-curricular volunteers are a completely unrelated (and self-selected) group. –jacobolus (t) 01:11, 11 July 2026 (UTC)

I don't think this is a problem about one particular topic per se. All of the Wiki-Ed students that I have seen come to technical (e.g. mathematical) articles have been similarly useless. At best the students write some half-assed essay-like text and dump it onto an article without any discussion or engagement where it either sits indefinitely, actively making the article worse, or eventually is removed or entirely rewritten. More commonly the students make a couple of auto-generated talk page spam sections and then never do any editing at all. I think the fundamental problem is that the teachers involved in these courses by and large don't seem to understand Wikipedia and don't make any apparent effort to engage with the Wikipedia community. In theory these could be valuable contributions and useful training/recruitment .... if the students tried asking for help and actually interacting with human Wikipedians. But since they don't, the entire exercise is basically a waste of Wikipedians time for no practical value. –jacobolus (t) 01:08, 11 July 2026 (UTC)

I want to correct some misunderstandings. Some of the comments in this discussion make it sound like WikiEd behaves as some sort of attractant, that causes instructors to start class projects on Wikipedia that those instructors would not have done otherwise. This is not true, and frankly, it's naive to think it might be true. To my knowledge, WikiEd does not advertise to educators to come to Wikipedia, and if the Wikimedia Foundation does any kind of outreach to educators, they would be doing so with or without WikiEd. Educators don't sit around wondering how to teach a class, and then suddenly get the idea to base the class on Wikipedia because someone at WikiEd contacted them and gave them the idea. I've worked for many years as a college professor, and I have some insight into how educators might get the idea of teaching through us. Wikipedia has a trendy appeal, and at some institutions it may look good to be able to say that you have a class here. And in some institutions, particularly those where there are low-paid instructors who are expected to process large numbers of tuition-paying students with as little friction and as much cash flow as possible, it can be attractive to think that if you just set your students loose on Wikipedia and let regular Wikipedia editors clean up any messes, that can be an easy way to teach a large class. That latter situation is where a lot of the disruptive class projects come from.

But WikiEd isn't acting as some kind of magnet to bring these classes here. They were coming here before WikiEd was created. And they'll keep coming even if WikiEd were disbanded, or banned from the English language Wikipedia. And it's useful to watch the reports of disruptive classes that come in to WP:ENB. Over time, one will notice (as I have) that the ones that cause editors headaches (and believe me, I'm not denying that student projects can sometimes be a real pain!) fall into two groups: the ones that never signed up with WikiEd, and the ones with instructors who ignored the instructions they got from WikiEd. Do away with WikiEd, and we'll still have just as many student projects here, but none of them will be supervised by anyone who understands Wikipedia. Thinking that if we eliminate WikiEd, our problems will be solved is just wishful thinking. It will actually make the problems worse. (And if you are one of those editors who gets headaches from student edits, please know that you are not alone, and be sure to read WP:NOTTA.) --Tryptofish (talk) 21:47, 11 July 2026 (UTC)

I think having students edit Wikipedia as a class project could be really valuable, especially for the students but also for Wikipedia. But it would take some amount of knowledge, engagement, and dedication on the part of their teacher(s), who need to be able to support/mentor the students (either directly, or possibly by recruiting community help). I haven't ever seen an example where that worked, and all of the teachers involved that I have personally come across seem completely disengaged. Has anyone had a positive experience with such student projects? –jacobolus (t) 21:59, 11 July 2026 (UTC)

Yes, I have, although it's the minority. If this interests you, follow WP:ENB. There was just discussion there yesterday about metrics on students in WikiEd projects who continued to edit after the class was over, and there are some who have stayed on and made thousands of good edits. --Tryptofish (talk) 22:08, 11 July 2026 (UTC)

Judging by the comments at Wikipedia:Education noticeboard#Should WikiEd be terminated? there are plenty of Wikipedians who are unimpressed with WikiEd. –jacobolus (t) 22:36, 11 July 2026 (UTC)

That's a single thread that gave rise to the one here. There are plenty of people on both "sides". As here, most of the "unimpressed" are also uninformed. --Tryptofish (talk) 23:06, 11 July 2026 (UTC)

I found the discussion at the WikiEd noticeboard after thinking about starting this discussion, not the other way around. I was looking for other editors who agreed with my complaints of Wiki Ed on Google and found the discussion. Ilov3gam3z (talk) 23:15, 11 July 2026 (UTC)

"Uninformed" seems pretty dismissive of people's complaints. As far as I can tell the supposedly "uninformed" "side" says "this seems like a mess and keeps causing trouble for us" and the other "side" says "you don't know what you are talking about and we can't do anything about the problems so too bad for you". Which frankly doesn't seem like a very solid argument. –jacobolus (t) 23:46, 11 July 2026 (UTC)

Wiki Ed does explicitly advertise itself and ask college faculty to consider teaching with Wikipedia. I see their messages a few times a year on one of the mailing lists to which I am subscribed (I'm an academic working at a US university). I don't know how often they solicit faculty participation, how widely they advertise, or how successful they are but they do engage in this activity and do not just wait for faculty to find them. I don't think this is especially important for the (misguided and incorrect, IMHO) accusations that motivated this entire discussion but I think we're all better off operating with more, correct knowledge. ElKevbo (talk) 00:46, 12 July 2026 (UTC)

Weighing in as someone who's repeatedly encountered but is not involved in WikiEd.

I've been around here for over 20 years. Prior to WikiEd, there were still instructors teaching courses. It was nearly as common back then for incoming instructors to be uninformed about how Wikipedia works. There was no way to find out about the existence of classes other than stumbling across edits on your watchlist. There was no way to find instructors ahead of time to try to educate them about Wikipedia. There was no way to ensure their classes included at least some material about reliable sources, NPOV, and encyclopedic tone before they let their students loose on the encyclopedia. Student edits were, if anything, a much worse problem.

Is WikiEd perfect? Of course not. Are there still lazy-ass instructors? Of course there are, and there will be with or without WikiEd. But now there's a structure that teaches them about Wikipedia and our policies and norms before they start their class. Now there's a standard curriculum that includes modules for learning about what constitutes a good or a problematic contribution to Wikipedia. And designating experienced editors as supporting wikipedia guides helps ensure there's someone available to mediate with the community when a mess is made, and help understand how to clean it up.

The average student edit nowadays is far less problematic than the average student edit 15 years ago. The really bad edits stand out, yes. But if you browse through the Articles tab of a few courses on the WikiEd dashboard, you're likely to also find large numbers of the small, productive, well-sourced edits that are easily overlooked; and a number of of-course-imperfect new articles on overlooked topics.

Yes, having to revert student edits now and then is annoying. We would have to deal with that with or without WikiEd. Without WikiEd, a far greater percentage of them would require reversion, and it would be harder to trace what other edits might need extra eyes.

I like to think of WikiEd as a safety rail on a steep road people could be driving on anyway. It's not the reason people drive there; and it doesn't prevent every crash. But it does make them, on average, less destructive. -- Avocado (talk) 13:13, 12 July 2026 (UTC)

A new page, Wikipedia:Mentorship noticeboard, has been created as a central discussion forum for mentors and editors interested in the Mentorship program. The noticeboard is intended primarily as a place for mentors to exchange ideas, discuss mentoring issues, share good practices, and suggest improvements to the Mentorship program itself. Mentors, prospective mentors, and other interested editors are invited to participate. Comments and suggestions about the new page are also welcome at the Talk page. Mathglot (talk) 10:27, 11 July 2026 (UTC)

Can anyone provide a link to a discussion about why the edit filter block action is not enabled on this wiki? Font8388608 (talk) 05:40, 12 July 2026 (UTC)

As far as I'm aware it's never been requested to be enabled here. See this Signpost story about the introduction of the edit filter (then known as the abuse filter) to Wikipedia, indicating that the introduction of automatic blocking would be delayed until the community had gotten used to the new feature. Graham87 (talk) 12:55, 12 July 2026 (UTC)

this configuration file, which I found from the AbuseFilter page on Meta, shows the wikis on which the abuse filter blocking action is enabled and indicates the Phabricator tasks where they were requested, demonstrating that blocking has to be enabled on a case-by-case basis for each wiki. Also see the Meta page on Requesting wiki configuration changes. Graham87 (talk) 13:15, 12 July 2026 (UTC)