The miscellaneous section of the village pump is used to post messages that do not fit into any other category. Please post on the policy, technical, or proposals sections when appropriate, or at the help desk for assistance. For general knowledge questions, please use the reference desk.
For questions about a wiki that is not the English Wikipedia, please post at m:Wikimedia Forum instead.
Discussions are automatically archived after remaining inactive for 8 days.
Latest comment: 1 day ago12 comments8 people in discussion
Hi all, I'm Nazneen, here for the Communications team at the Wikimedia Foundation. Back in May, we ran a pilot campaign encouraging mobile web readers to download the Wikipedia app, as part of the 25th anniversary work. Thank you to everyone who weighed in on that. I wanted to share how it went and what we're planning next.
In the pilot, there were around 701,000 app installs. That's a strong signal that there's real interest among Wikipedia readers to install the app once they know about it. This lines up with what the Apps team shared in April about why we think the apps could be important in the future: as more people find information through AI summaries instead of visiting Wikipedia directly, we want readers to make Wikipedia an important part of their knowledge diet. The apps are a place where readers can build their relationship with Wikipedia, receive push notifications, interact with a home screen – features that readers love that we can't give in a browser. The thing is (and we’ve seen this a lot in the comments on our social media posts) that when Wikipedia readers find out about our mobile apps, they are really excited – but they just don’t know about them.
Given that response to the campaign, we would like to keep this going. Rather than a single pilot, we're planning an ongoing, lower-key presence encouraging app downloads across the year ahead.
We'll begin with the same simple banner that performed best in May's pilot, so it's a continuation of what already worked, that can be adjusted based on performance and how readers respond. There may be a small number of short periods later in the year where it runs more intensively to test impact, but we'll flag it here if there's a meaningful change to the overall approach.
We know that some other platforms push their mobile apps so hard that it makes their mobile websites difficult to use. We definitely don’t want that – people should be able to happily get the goodness of Wikipedia from the web browser. This plan is about making sure readers who'd genuinely value the app know it exists; it's not a move toward restricting or diminishing the mobile web experience. As with the pilot, this will only show to logged-out readers on the mobile web, for each reader it will be capped at 6 impressions per month and if you dismiss the banner, you won't see it again on that device that month.
We'll also make sure these plans don't crowd out any planned community banner campaigns, adjusting our own scheduling as needed to give those the space they require.
Given the push for mobile users to use the app, I would love it if they could be granted access to the entire platform. As a mobile user I find it really frustrating that I can’t search for categories, or that the search is limited to a finite number of results. ExtantRotations (talk) 15:27, 19 August 2026 (UTC)Reply
I am Amal Ramadan and I support Wikipedia apps team, for search the categories:
If you type Category:<category you want>, you should be able to find the categories you are looking for. Regarding generating a longer list of search results and improving search outside mainspace, at the moment our team is focused on getting semantic search to work well on Wikipedia. Once we have that, we can do much more, and you are welcome to subscribe to the apps newsletter for the full news and updates of the apps' work. ARamadan-WMF (talk) 17:53, 20 August 2026 (UTC)Reply
How about adding more support for editing in app? The mobile web editing experience is problematic as it is, and I was extremely disappointed to find out the app has even less functionality. I have no issues simply reading articles in my mobile browser and I see no added benefit to clogging up my device with yet another app to do the same thing, so I promptly uninstalled it. ChompyTheGogoat (talk) 02:39, 25 August 2026 (UTC)Reply
I'd have to go back and install it again to figure out what exactly is missing or broken. I'd like to know what the thought process was behind releasing a lower functioning app in the first place. What purpose does it serve? ChompyTheGogoat (talk) 22:52, 27 August 2026 (UTC)Reply
Hi @ChompyTheGogoat my name is Jaz, I am a Product Manager on the Mobile Apps Team.
We agree that there is room for improvement with the editing experience in the apps. To make it better, we're currently actively working on a feature to allow people who edit from the app to access the VisualEditor, and we could use your input. And if there are other editing features you'd like to see on the apps, I'm happy to chat about them.
The web still plays a very important role and is the front door to welcoming in the diversity of reader interest. With that said, I do want to be realistic about how we see the apps. We are investing in them primarily as a home for our most engaged readers, and our research indicates that they appreciate features like Year in Review that we can provide only on the apps. The apps are our highest retaining space for readers, which is important when Wikipedia is experiencing declines in pageviews. These highly-retained readers are good candidates to become future editors, so we want them to have a positive experience when they are ready to become editors, hence the attention to meaningful handoffs to desktop/mobile web, where the superior editing experience lives, and where we are focusing all of our effort to continue to improve the desktop/mobile web editing experience. I hope that provides a bit of insight into our thinking. I also welcome you to check out this video from Wikimania which goes into greater detail around this thinking. JTanner (WMF) (talk) 15:49, 28 August 2026 (UTC)Reply
I think that line of reasoning is problematic. You want people who are invested in Wikipedia to download an app, become more invested, find out the app isn't useful for editing, and uninstall it to go back to the website? Or keep swapping back and forth between the app and the website every time they run across something they want to edit? Something like Year in Review is a minor bonus feature, not major functionality. It's the kind of thing commercial corporations use to drive sales while discouraging people from navigating away (and it's a gimmicky fad that will probably die down in a few years). This just furthers my distaste for WMF pursuing the same kinds of tactics. Our purpose is to be educational, not obsessively chase engagement stats. Who cares if someone navigates away to continue learning about the topic somewhere else instead of falling down a wiki rabbit hole for hours? As long as people consider it a reliable source of information that they utilize regularly we're accomplishing that goal. Websites should provide clean basic functionality for their primary purpose, while apps should offer an expanded range of options for more dedicated users via more complex programming that can't be routed through browsers well (especially on mobile). Like all the customizations and workarounds that are currently accessible through a hodgepodge of scripts, gadgets, and beta features, which sometimes interfere and cause glitches but overall improve my experience compared to the default native environment that's already glitchy and not optimized for mobile. The problem I tend to see these days is when programmers turn around and try to stuff that functionality back into a website, and then they don't care if it breaks because their ultimate goal is to funnel everyone to the app anyway. Some sites don't even functional well on my desktop anymore because they've gotten so ridiculously bulky. Here we have the opposite problem where you've distilled it down to basic functionality for a slimmer app, then hung a few bells on to try to make it appealing. It's very bizarre and once again feels more like "keeping up with the Joneses" than a truly introspective view for how an app would support the overall purpose of Wikipedia. If the people who spend the most time on the platform and are the ones who actually maintain it aren't utilizing the app, that speaks volumes for its actual usefulness. ChompyTheGogoat (talk) 03:21, 29 August 2026 (UTC)Reply
If it is expected that users hop between platforms, that makes it all the more baffling that notification dismissal isn’t shared between them. Why make users acknowledge each notification twice? ExtantRotations (talk) 16:07, 29 August 2026 (UTC)Reply
@NNawaz-WMF, @ARamadan-WMF, @JTanner (WMF) I usually read and edit Wikipedia on a tablet, using a browser. The browser interface and the editing tools are fine for me (I use source editing). I use Wikipedia very occasionally on a phone or a desktop pc.
How is the app better? As I said, the Web interface seems great. I don't understand the need for an app. Many apps are useless, in my opinion.
Why does Home Depot, for example, even have an app? Their Web page works just fine on a phone or a tablet, and their app gives no advantages. Many other companies with perfectly good Web pages also have apps... which seems (to me) like useless effort.
One thing the app desperately needs is support for namespaces outside of Articles and Talk pages. For everything else, it just brings you to a web interface. At the moment, I see the customizability and extra tools on mobile web much more useful than the app. Axolitl(talk|contribs)05:27, 31 August 2026 (UTC)Reply
CentralNotice for Bangla Wikivoyage Article Contest 2026
Latest comment: 4 days ago9 comments6 people in discussion
A contest will take place from September 1, 2026, to September 30, 2026, on Bangla Wikivoyage to encourage the creation of new articles and improvement of existing content. A CentralNotice request has been placed to promote the contest to users of English and Bangla Wikipedia, including non-registered users from Bangladesh and the Indian state of West Bengal. Thank you. ARI (talk) 14:26, 22 August 2026 (UTC)Reply
@ARI what would be the message on the CentralNotice banner for enwiki that would avoid an element of surprise for the English readers if they are clicking through, given that it goes to straight to Bangla language project? Also the CN link here is not valid? – robertsky (talk) 17:36, 22 August 2026 (UTC)Reply
I oppose showing more than one banner a week, running these banners for unregistered readers, or directing enwiki viewers to pages not written in English. Banners are not free. Best, HouseBlaster (talk • he/they)16:59, 23 August 2026 (UTC)Reply
Well, that's just not going to happen. The communities have more than 52 banners they want to run each year. Furthermore, many banners are only shown to people in certain countries, at certain wikis, or using certain languages, so even if we run multiple banners at the same time, one person might see none, another two, and someone else three. Speaking of three, many banners are set to only show three times (=three pages) in a week. You've recently been editing an average of ~30 pages per day, so even if you saw all three of the banners in my example, you'd still only see a banner on significantly less than 5% of your page views. And if you'd like to make that be zero, then let me suggest spending a moment in Special:Preferences#mw-prefsection-centralnotice-banners. WhatamIdoing (talk) 15:48, 27 August 2026 (UTC)Reply
(CN admin comment) @HouseBlaster @Robertsky The banner is targeted specifically to Bangladesh and West Bengal (India), rather than to English Wikipedia readers globally. This targeting has also been used for previous Bangla Wikivoyage Article Contest campaigns, including in 2024 and 2025.
Given the geographic targeting and the fact that the campaign is intended to reach potential Bangla Wikivoyage contributors in these regions, do you still have concerns about showing the banner with these settings? —Yahya (talk • contribs.) 18:27, 27 August 2026 (UTC)Reply
I do still have the same concerns, yes, even though the banners are geo-targeted. I would've opposed the terms of the 2024 and 2025 campaigns if I had been paying attention. Best, HouseBlaster (talk • he/they)20:04, 27 August 2026 (UTC)Reply
@HouseBlaster, @Robertsky, @Omphalographer It is my first time applying for a CentralNotice, I apparently made a poor assumption about how the banner would be perceived by English Wikipedia readers. I thought users familiar with Bangla would understand that it was leading to Bangla Wikivoyage. Also, the link in my original post was incorrect; the intended link is the CentralNotice request page mentioned above. My apologies for the confusion. ARI (talk) 03:00, 25 August 2026 (UTC)Reply
This was apparently tagged as a result of a massive RFC (shortcut: WP:LUGSTUBS2) three years ago. I have barely begun to read through the RFC but it appears that this article meets the criteria for draftification. It's not clear to me why this hasn't been moved to draft space yet. If there's a reason this was spared, and it's not just an oversight or a result of a large backlog, then I would expect that to be clearly documented. Pinging the RFC closer @HJ Mitchell, who might have some insight into what happened and should happen here. —Myceteae🍄🟫 (talk) 17:11, 22 August 2026 (UTC)Reply
What happened is that a lot of editors voted for someone else to do the work, and then got mad that other WP:VOLUNTEERS didn't instantly do their bidding. Voting for someone else to do work that I refuse to do myself has been one of the more unfortunate trends over the last few years. (For clarity: the OP is not one of these people. The OP is only asking about a confusing message on an article.)
Then it turned out that a lot of these articles either shouldn't be moved to the draftspace at all, because they're notable athletes, and nearly all of the rest shouldn't be moved to the draftspace because they should just be redirected to a team roster. More than 95% of the original list has already been handled; just 54 tagged articles remain in the mainspace today. Most of them are in South Asian countries. Wikipedia talk:WikiProject Cricket is the place to post if you want something useful done about this one. WhatamIdoing (talk) 15:58, 27 August 2026 (UTC)Reply
Latest comment: 2 days ago3 comments3 people in discussion
I see a lot of dating using the religious AD and BC while others use the modern secular BCE and CE. As this platform is used by more than Christian readers, I propose standardized CE and BCE throughout. ~2026-46089-12 (talk) 05:47, 23 August 2026 (UTC)Reply
See WP:ERA. Wikipedia is built by people from all over the world, and with many different backgrounds. Therefore, standards for certain issues (BC/BCE, formats of dates, citation style, color/colour, and probably more) are not imposed on everyone. Johnuniq (talk) 06:21, 23 August 2026 (UTC)Reply
Latest comment: 2 hours ago116 comments17 people in discussion
Please for the love of all that is holy, can we either revamp these to require some degree of human oversight in terms of suggestions, or just remove them entirely? My experience both with peeking at them briefly as a newbie myself and reviewing edits made by others is that they give vague and unhelpful suggestions on articles that need major improvements, so we get these new editors with no understanding of the subject making equally unhelpful (and sometimes nonsensical) changes that other editors then have to review and usually revert. It's creating extra work without actually improving the articles. It would be one thing if the tool was actually capable of identifying truly minor edits needed, like spelling and grammar, but that doesn't seem to be the case. We'd be better off leaving newbies to look around on their own and make edits where they feel they adequately understand both the subject and the changes that are needed, rather than trying to guess what some automation has identified. ChompyTheGogoat (talk) 02:31, 25 August 2026 (UTC)Reply
I'll just dump here what I wrote on my user page a while ago:
"I started editing by going through "Newcomer Tasks". After three days and a handful of edits, I have given up. The vast majority of the "Newcomer Tasks" I was shown were completely unsuited for actual newcomers. I mostly looked at tasks in the history topic. Most of them consist of being asked to improve very poor articles on very obscure topics. In many cases, there seem to be only few books or articles covering the article topic, and they tend to only be available in specialized libraries.
I do not know how these "Newcomer Tasks" are chosen. Based on what I've seen, most of the articles seem to have what I found to be called "maintenance templates". I suspect that these articles are automatically assigned to newcomers based on these templates.
That is a very poor way of introducing newcomers to editing Wikipedia. It feels more like being given the odious tasks that nobody else wants to perform, than tasks tailored to the needs and skills of newcomers. It feels like starting an internship and being given the tasks of making coffee and cleaning up behind the staff.
I expect those "Newcomer Tasks" to drive away a lot of people who might have gone on to become valuable contributors if they hadn't been turned off right at the start. While that isn't the case for me, it seems that I will have to find my own way on Wikipedia." Long is the way (talk) 13:44, 26 August 2026 (UTC)Reply
At least those (IRL) tasks are actually easy to understand, just tedious. These are more like being asked to tidy up and walking into a building actively on fire. The ones I've seen newbies attempting to "fix" lately didn't have maintenance templates that I recall, so I really have no idea how they're being suggested. They suddenly started popping up on a group of related articles that have received attention recently. ChompyTheGogoat (talk) 17:12, 26 August 2026 (UTC)Reply
Nope, no template. The latest was a "revise tone" with the summary "Removed bias and slang", when it fact the only substantial change was a misinterpretation based on lack of knowledge of the subject matter. There was no slang of any kind. ChompyTheGogoat (talk) 17:46, 26 August 2026 (UTC)Reply
Before I settled on the History topic area I also looked at tasks in the Philosophy and Religion area and the Computer (it's been a while, I'm not sure if that was the title) area. In the Philosophy and Religion area, most articles that showed up were about random churches, religious colleges and religious schools in the US. And if that's not bad enough, every time I did a quick search for reliable sources (I won't do a thorough search for an article about a random church that interests me not one jot) came up empty. So what should I do? Nominate article after article for deletion as a newcomer when the author couldn't be bothered to provide proper sources for their edits and someone else couldn't be bothered to do more than tag it? And in the Computer area most articles were about random software and tagged for promotional language. Again, a quick search for proper sources usually came up empty. Once you've wasted a few hours like that it's hard not to come to the conclusion that Newcomer Tasks (if not Wikipedia in its entirety) are thoroughly broken. Long is the way (talk) 18:21, 26 August 2026 (UTC)Reply
I think I was looking at Science - which is incredibly broad, with no way to narrow down my actual interests - and my "1-2 minute copyedit" tasks needed a full top to bottom rewrite (and probably sourcing too). I literally didn't know where to start. Instead I just dabbled with very minor fixes that I stumbled across during my normal reading, or occasionally when someone else mentioned it at Teahouse and didn't know how to do it themselves. Honestly, I think basing it on templates would be better than how it currently is. Even moreso if there was a specific "newcomer level" that could be appended to indicate that it's an easy fix for someone who's still learning - based on actual human judgement - leaving the more complex issues out of the newcomer database. Of course most experienced editors would just fix something that simple, but there could be an active choice to leave it in if it doesn't interfere with the overall reader experience. ChompyTheGogoat (talk) 19:22, 26 August 2026 (UTC)Reply
Thank you for taking the time to describe your experiences. I work with the Growth team, which developed newcomer tasks, so I wanted to let you know that I'm following along and thinking about all of the feedback shared in this thread.
New editors make mistakes, and that's true whether they arrive through Suggested Edits or on their own; our goal is to make those early mistakes smaller and easier to learn from, and we know we don't always succeed. Suggested Edits clearly aren't the right path for everyone, but in multiple controlled experiments they've increased the share of newcomers who make a first edit and who are still editing weeks later (2020 analysis, 2021 analysis, 2025 analysis), so many new account holders do find them valuable.
Two current efforts speak directly to what you've described. Newer tasks like Revise Tone are far more in-context than the broad "copyedit this article" tasks you encountered: they point to specific sentences rather than leaving a newcomer staring at an article that needs a rewrite. ChompyTheGogoat, since your recent example was a Revise Tone edit gone wrong, I'm curious if you've looked at other Revise Tone edits? Although newcomers are still making some mistakes, it seems like this task is helping provide enough structure to support newer editors, while still teaching more valuable skills than a super simple task like Add a Link.
And our Early onboarding / Home experiment is testing whether newcomers do better when they can choose specific interests instead of overly broad topics like "Science", which is exactly the gap several of you have identified. In early testing, this surfaces far more specific suggestions and lets people find the niche topics they actually know something about. Does that sound promising?
Finally, as @Johannnes89 notes, much of this is tunable locally via Community Configuration (which tasks are enabled, which templates feed them).
Like I said, there needs to be actual human oversight for me to consider it viable - not problematic automation, and absolutely not this LLM suggestion crap. We spend far too much of our time fighting LLM content and in no way shape or form do we need it further misleading new editors. The problem isn't just the edits they make using such a tool, but what they would learn from it and apply in the future. And on that note, just the fact that newcomers are more likely to keep editing is not a very useful metric on its own - the edits themselves need to be beneficial. Quality over quantity. How many of the newcomer task edits are actually reviewed by more experienced editors, and what's the reversion rate on them compared to non-suggested newbie edits? How many of those editors continue to be productive over months or years, not just weeks? The automation we need is education - a walkthrough for new editors that covers the basics of editing; both the technical how-to aspect and an overview of the four pillars plus the most crucial guidelines. WP:NOTABILITY, WP:COI, and WP:NOLLM come to mind as the most obvious "read before editing" candidates (with checkboxes for the latter two affirming whether or not they have a COI to disclose up front, and that they agree to not insert LLM content). Actually teach people what they should be doing instead of just tossing them in the deep end and saying "here, change something and wait to see whether you get corrected or not". WikiEdu has been highly successful, right? There's no reason we can't package up the basics for all new editors who don't have time or access to such programs. Add an FAQ in too, and links to various sources for additional help.And as long as I'm ranting - better navigational structure. A site index. If I'm wondering "is there a guideline about this" or "where should I report such and such problem" I should be able to scan a list for anything that sounds relevant, instead of attempting to dig through namespace filtered search results - and newbies may not even know about namespaces yet. Once again, more often than not it comes down to "screw up and get corrected", which is a valid learning method but shouldn't be the primary one, because it gets extremely discouraging. Far better to set people up to succeed. ChompyTheGogoat (talk) 21:20, 26 August 2026 (UTC)Reply
You raise a fair point about quality over quantity, and it's one the team shares. All of the Growth team's experiments account for reverts: when we report that Suggested Edits increase activation and retention, we're counting only "constructive" activation and constructive edits, meaning edits that were not reverted. A newcomer whose changes get reverted isn't counted as a success in that data (definitions in our data glossary). If it would be useful, we can also pull recent English Wikipedia data comparing revert rates on Newcomer Task edits vs. other newcomer edits; just say the word.
I've also filed phab:T436196 asking Movement Communications to share more about newcomer metrics as a whole, because I suspect some of the current frustration may relate to the recent increase in new accounts and new editors. More newcomers means more newcomer mistakes reaching patrollers, even if per-editor quality hasn't changed. Does that match what you're seeing?
On education: this is an active area of work. Together with the Community Development team, we're developing short micro-learning videos based on the Wikimedia Core Curriculum, to help new contributors understand the basics and build confidence before and while they edit. That said, no single onboarding path works for everyone: some people want to read the guidelines first, some learn best from a video walkthrough, and some only absorb things by trying a small edit and getting feedback. If we want an encyclopedia written by a broad, representative group of editors, and one that stays as neutral as possible, we need to support several ways in rather than optimizing for just one type of potential editor.
Where I fully agree with you is that we can do more to set new editors up to succeed, and your navigation ideas are a good example. What would a site index that doesn't overwhelm a newcomer look like to you? Namespaces alone are a bizarre concept for most newcomers to grasp, and we do very little to explain them. As always there's so much room for improvement and limited capacity to "make it so" but I'm committed to doing the best I can to improve onboarding for newcomers on the wikis. Thanks - KStoller-WMF (talk) 00:16, 27 August 2026 (UTC)Reply
Yes, I would like to see the comparison stats. I'm particularly interested in which ones have actually been reviewed and confirmed to be an improvement vs just not noticed, but I realize that's harder to prove. I'm aware that new editors make lots of mistakes, but my personal experience has been that nearly 100% of those flagged as newcomer tasks are at best useless, if they don't actually make things worse, vs more of a dice roll for normal newbie edits. What exactly are the existing criteria for them to be suggested anyway?As far as the education side, I'm someone who prefers text learning over videos, so I'm aware that multiple approaches are needed. What I'm suggesting here is just a very brief intro - a popup (with option to skip) with a few slides mentioning the absolute basics and linking to additional learning resources. <2 minutes to get through. I would hope anyone who wants to edit Wikipedia could handle that amount of text, and the COI/LLM agreements are universal. As mentioned elsewhere, the situations we want to avoid the most are the novice good faith editors who are truly WP:HERE and end up with significant reversions purely because they're unaware. We had a case at WP:AINB recently where a new editor had made substantial LLM changes across numerous articles before anyone noticed and called it out. They were very apologetic and actively participated to help with the cleanup. Those are editors with real potential that we don't want to discourage. And removing plausible deniability would streamline disciplinary actions on other cases as well.For navigation, at minimum we should have top level links in the main menu to WP:List of policies, WP:List of guidelines, WP:Manual of Style, and WP:Noticeboards. Maybe WP:Template index too. I'd also recommend considering a customizable shortcuts section that logged in editors can add any pages they want to reach quickly to; internal bookmarks. And personally I dislike the current structure of internal navboxes such as Template:Wikipedia policies and guidelines - in theory if I'm already on the right overall section I can jump around from there, but my brain has a tendency to skip over them because they feel disorganized, and it's worse the busier they get, like with that example. I think stylistic changes could help with that - maybe color coding, collapsible sections, and some kind of change to the layout of the lists themselves within the cells? It's not something I've given much consideration to, but could be workshopped here at VP. And for the sake of being thorough, there could also be one site directory page linked in the footer with a complete list of all the main internal pages in WP space. Obviously It can't be 100% comprehensive, what with all the subpages and minor pages that are frequently created or deleted, but primary perennial ones. Other projects could implement these ideas too - if I want to go edit on one I'm unfamiliar with it would be helpful to know I can go through the menu to find their own policies and guidelines to ensure I'm in compliance with any standards that are different from here. It's especially difficult to try to track such things down via searches if you're dealing with foreign languages (I swapped out a few images across several foreign wikis the other day to avoid breaking pages via changes made at Commons).I'm probably well over my allotted time here, and it's only tangentially related, but one other idea I had recently was some area for more collaborative work on article creation. Not just the brief feedback from reviewers or general "how to use Wikipedia" questions for mentors, but a longer term partnership aimed at getting articles completed and published together. I'm sure newbies would find it the most useful, but I can also foresee situations where people need a specific type of help - for example, one person might be a subject matter expert while the other can help with translation. It could potentially improve AfC rates as well as the initial quality of articles that are directly published in "notable but needs work" condition. ChompyTheGogoat (talk) 05:11, 27 August 2026 (UTC)Reply
Example pop-up for the "Find references" task
What I'm suggesting here is just a very brief intro - a popup (with option to skip) with a few slides mentioning the absolute basics – that's exactly what each newcomer task offers? Johannnes89 (talk) 05:47, 27 August 2026 (UTC)Reply
Could you reduce "how to edit Wikipedia" to a handful of sentences? In my experience, "how to" depends a lot on the context, and what you're trying to accomplish. How to fix poop vandalism is a completely different skillset from how to add a new paragraph. WhatamIdoing (talk) 17:35, 28 August 2026 (UTC)Reply
Of course - I just mean the very basics that would be most useful to good faith editors with zero experience. Brief references to things like WP: NOTABILITY, WP: RELIABLE SOURCES, WP:NPOV, and WP:MOS as well as the aforementioned WP:COI and WP:NOLLM, with links to all of these places as well as additional resources like WP:TEAHOUSE and WP:HELPDESK. We can't stop vandals from being vandals, and we can't fit all of the educational material into a single popup, but we can inform people that these things exist and help them find them, to hopefully prevent some of the most common genuine mistakes. I couldn't begin to count the number of new editors who come to Teahouse asking about a reversion, warning template, etc based on guidelines that they had no clue even exist, because how would they? Even the welcome templates are only dropped after they make an edit, see the notification, and go read it. We should be offering these resources upon account creation/first attempt to edit to be proactive instead of reactive. We could also include a mention of reversions and why they're a part of the learning experience (even for seasoned editors), not automatically criticism, to help people feel less offended by it. ChompyTheGogoat (talk) 03:35, 29 August 2026 (UTC)Reply
Most newcomers don't try to start an article, so why should they care about our notability rules? Similarly, MOS violations are usually easy enough for editors to fix, and it's thousands of small rules, most of which are either automatic (basic grammar) or irrelevant (e.g., the name of a gene should be italicized, which 99.9% of newbies will never need to know). I wouldn't bother with that. But RS and NPOV and COI and NOLLM all sound like reasonable things for us to educate people about. WhatamIdoing (talk) 04:13, 29 August 2026 (UTC)Reply
I think the definition of "most" is debatable, but certainly so is the specific content that should be included. I'm just trying to get the overall concept across. Which mistakes are good faith new editors most likely to make in their first handful of edits, and what can we offer them that would be the most useful to help avoid those? Collecting a pool of early edits that have been selected for good faith attempts at improvement - filtering out vandalism etc - would help us establish specific targets, and there would probably be some adjustments as we see the results. ChompyTheGogoat (talk) 04:21, 29 August 2026 (UTC)Reply
The last time I saw the numbers, which was some years ago, about 25% of newcomers tried to start and article. Therefore, 75% of newcomers didn't. 75% is "most" under all mathematically sound definitions.
Learning by doing, especially learning from mistakes and feedback, is very effective when the experience is not demoralizing. The challenge is to keep this in balance. And getting corrected, or reverted, is unavoidable.
Newcomer tasks are visible and easy to categorize. This permits testing, tracking, improvement. It can also promote confirmation bias and a (possibly false) sense among experienced editors that newcomer tasks are especially error-producing. Newcomers who get "bitten" for completing a structured learning activity understandably feel let down. Even if the net success rate of these tasks is better it is for newbies left to their own devices the experience can be frustrating.
Related to (1) is figuring out the optimal way to introduce our myriad policies, guidelines, practices, and jargon. Presenting too much "required reading" up front is likely to discourage some newcomers while others feel set up for failure by not having fundamental principles put in front of them. We should make this information visible and accessible in a variety of ways. There will still be problems. I like policies and guidelines but they have to be applied and interpreted in context.Related to (2), it's funny that you mention WikiEdu. My sense is that it is successful but it is a frequent topic of discussion. Some editors feel that it disproportionately generates bad contributions that require cleanup. I haven't seen convincing evidence of that but WikEdu contributions leave a trail and (may) come with a set of expectations, like newcomer tasks, that can increase the frustration. These are good problems to talk about, and it's beneficial to have relatively new editors in these discussions. —Myceteae🍄🟫 (talk) 01:18, 27 August 2026 (UTC)Reply
See my above comment re: 1. My intent for that is just a very brief "Welcome to Wikipedia, here are a few of the most crucial things you should know and places you can go to learn more", not a master's course in editing.I don't have a lot of experience with the outcome of WikiEdu myself - I'm mostly going off what I've seen others say - but what little I have seen usually seems to fall more in the "this is a good start, but here are some more suggestions" where reverting and explaining is helpful, as opposed to "this never should have been suggested in the first place so there's nothing to improve and both of our times were wasted". I definitely could have benefited from more useful suggestions; I was very wary of making any mainspace edits until quite recently and stuck to extremely simple ones (the opposite reaction from LITW). My suspicion - again difficult to prove - is that there's an inverse relationship, where those of us who have a better understanding of Wikipedia from the start and are able to handle the learning curve better look at these tasks and see what's inherently wrong with them, while those who don't know how anything works here assume the suggestions are valid so they just go ahead with changes even when they don't really understand the assignment. ChompyTheGogoat (talk) 05:32, 27 August 2026 (UTC)Reply
The other big differences with WikiEdu is that there are much fewer edits produced by the program, and that when people have pointed out issues with those edits, they actually did make changes to the workflow that seem to have improved the situation. Gnomingstuff (talk) 18:05, 28 August 2026 (UTC)Reply
Yeah, I'm sure it's not perfect, but right now it's our best resource for a more structured program to help support new editors, so I think we should be using what's been learned there to do the same in a more hands off way for others who don't have access to such things. Use what works and discard what doesn't. ChompyTheGogoat (talk) 03:41, 29 August 2026 (UTC)Reply
@ChompyTheGogoat, what does "actual human oversight" mean to you? From where I'm sitting, the newcomers are human, and so if and how they decide to make the edit constitutes "actual human oversight" of the edit already. But I think you mean something else. WhatamIdoing (talk) 16:08, 27 August 2026 (UTC)Reply
Oversight by someone with more experience (hopefully) of what actually gets added to the task database, to ensure the suggestion itself is valid and easy to comprehend. ChompyTheGogoat (talk) 22:49, 27 August 2026 (UTC)Reply
Are you volunteering to check all the pages that are identified as needing work, to make sure that they actually need that kind of work?
We need about a thousand brand-new accounts to not only register, but also to make their first edit every day. Anything that reduces that number risks Wikipedia's future, because I am going to die. Only a small fraction of them will complete a Newcomer task, but the newbies who do those tasks usually do multiple edits to multiple articles. We probably get about 1,000 to 1,500 newcomer tasks completed per day. Some tasks result in multiple edits to the same article, but we also need a buffer in case a pre-screened article doesn't find an interested editor. I estimate that manually pre-screening would therefore require pre-screening about a thousand articles a day. At a sustained rate of one article per minute, that's 16 hours of work, every single day of the year. It would also have the downside of introducing personal preferences (e.g., this editor wants to minimize links, that editor is unusually sensitive to 'promotional' content...). Do you think it would be worth it, in terms of improving the edits? WhatamIdoing (talk) 17:07, 28 August 2026 (UTC)Reply
If only a small fraction of new edits are made through newcomer tasks then yes, I absolutely agree it would be justified to spend more editor time screening tasks instead of going back and fixing bad edits that result from them. Higher quality suggestions would also result in higher uptake by those of us who avoided them because they're problematic. Even if they actually were based on templates, as has been suggested in this conversation but not substantiated by the evidence, that would mean a human editor read the article and chose to add the template - something that already happens and doesn't add labor. Turns out my gut was right and this is a BS LLM doing BS LLM things. Sure took some tooth pulling to get that admitted. ChompyTheGogoat (talk) 04:15, 29 August 2026 (UTC)Reply
My question isn't whether you think somebody else should prescreen the articles for each task. My question is whether you wanted to do that.
Different tasks have different triggers. The tasks also change over time. For example, the Add a link task used to look (only) for Template:Underlinked; now it is based on a statistical calculation.
Related to this, can we please disable the link suggestions feature in mathematics articles? It consistently causes new editors to add links which are either overlinking or even semantically incorrect (i.e. a different concept with the same name). These editors are often not mathematically advanced enough to understand the difference between a good link and a bad link in a mathematical article. It just ends up creating work for others who have to revert these changes. Elestrophe (talk) 00:05, 28 August 2026 (UTC)Reply
Isn't that one of the newcomer tasks also? Same overall issue. They mean well, but the suggestions just aren't good for newbies with no Wikipedia experience or subject matter knowledge. ChompyTheGogoat (talk) 01:28, 28 August 2026 (UTC)Reply
It is a newcomer task. There was a discussion about this quite recently: Wikipedia:Village pump (proposals)/Archive 231#We need to get rid of the "suggested links" tool. Some tweaks were made and other potential interventions suggested or were already being worked on that might improve the fidelity. There's a lot of discussion there of data indicating that links created via the newcomer task get reverted less often than links inserted by newbies going at it alone. There were some questions about the precision of these figures but nothing that suggested to me that the observation was directionally wrong. —Myceteae🍄🟫 (talk) 01:52, 28 August 2026 (UTC)Reply
That statistic doesn't mean the feature is a good thing. It's not like the existence of the feature prevents new editors from adding links they would have already made, so it's still just creating a bunch of bad links that have to be reverted. Elestrophe (talk) 02:58, 28 August 2026 (UTC)Reply
You're correct that the existence of the tool doesn't prevent manual edits, but the numbers show that it does encourage newbies to make that kind of edit. If nothing else, it educates them that this is the kind of thing that Wikipedia wants to have done. WhatamIdoing (talk) 17:12, 28 August 2026 (UTC)Reply
@Myceteae Re "links created via the newcomer task get reverted less often than links inserted by newbies going at it alone", is there? I've seen data on add a link vs overall newcomer edits, but not specifically vs non-task link additions. CMD (talk) 04:21, 29 August 2026 (UTC)Reply
I count 94 edits to the mainspace, of which a total of 50 were newcomer tasks and a total of 63 were reverted. Specifically, I count 34 reversions of newcomer tasks (68% of newcomer tasks) and 29 reversions of ordinary edits (66% of non-newcomer tasks).
Maybe you want to run some proper calculations, but that doesn't sound like a statistically significant difference to me. Therefore, I think it would be difficult to blame the existence of newcomer tasks for those edits. WhatamIdoing (talk) 17:22, 28 August 2026 (UTC)Reply
The reason all of these edits have not been reverted is because I have not slogged my way that far down the list yet, and because several of the edits have been subsequently buried under a deluge of other edits making cleanup even harder.
No, the reason all of these edits have not been reverted is because the community did not consider them worth reverting. For example, picking a "revise tone" example from the middle of their contribs, I see this:
Perhaps the greatest Harbor Dynasty was that of Girls' Soccer who won 9 CCS Championships in 14 years → Harbor High School has seen notable athletic success over the years. The Girls' Soccer team won 9 CCS Championships in 14 years
It may not be perfect (e.g. neither version complies with MOS:SPELL9), but I think that rephrasing it to get rid of puffy "greatest Harbor Dynasty" language constitutes an incremental improvement. Maybe you would agree with me.
Remember that you don't have to clean up after a thousand newbies each day all by all yourself. In fact, you don't have to do any of it, unless you actually want to. WhatamIdoing (talk) 20:58, 28 August 2026 (UTC)Reply
No, the reason all of these edits have not been reverted is because the community did not consider them worth reverting.
But you still have to go through every single one to see whether they are or not. Many of them are.
I do not actually "clean up after a thousand newbies every day all by yourself," because there are not enough hours in the day for that. I don't see why I am the one who is being scolded here instead of the people creating the cleanup work. Gnomingstuff (talk) 22:11, 28 August 2026 (UTC)Reply
I picked one out randomly from the user's contribs, and I found no need for it to be fixed. Why should I assume that all the others need fixing, or even most of them? It's of course possible to find the one outlier, but it's not generally reasonable to assume that the one you found is an outlier.
I checked another, chosen for being a net negative number of bytes (because I thought a reduction in page size would be less likely to be a whole-page re-write, and I didn't feel like looking at a complex diff). That, too, was a good edit – not perfect, but better than what was there before.
My point isn't that the editor is any good. My point is that you don't have to take on reviewing those edits unless you actually want to. Those edits have almost certainly been reviewed by someone else. That someone else will be less adept at your particular skills (we are all less adept at AI detection than you), but they will have been checked for an ordinary level of reasonableness, and determined not to be obviously bad. As a result, it's IMO not necessary to treat this user's contribs as a significant threat to Wikipedia. Review them if you want, and don't if you don't. WhatamIdoing (talk) 20:00, 30 August 2026 (UTC)Reply
It doesn't indicate that they improve edits at all either. That's exactly the point I was getting at all far as editors who use newcomer tasks at all vs the total sum of new editors, which includes vandals, UPE, SPAs and the whole range of LTA sockers. Most people who come in with any form of bad faith won't bother with the tasks, unless it's purely an attempt to game user rights. ChompyTheGogoat (talk) 03:50, 29 August 2026 (UTC)Reply
This person seems to be a WP:SPA. An unorthodox one, to be sure, but their editing pattern is the same: spam out a bunch of newcomer tasks until Number Goes Up enough that they can do what they're really here for. In this case that's a low-quality draft on their favorite math problem rather than a low-quality draft on their marketing startup, but the pattern is the same. They even all but admit here that they mostly care about getting their edit count high enough to get permissions. Gnomingstuff (talk) 18:11, 28 August 2026 (UTC)Reply
It's once again not clear that the newcomer task creates or promotes the problem as opposed to being a thing that can be used in conjunction with extremely common problematic newbie behavior. There are always newbies who try to juice their numbers so they can gain more tools and start working on their pet projects with fewer restrictions. —Myceteae🍄🟫 (talk) 21:09, 28 August 2026 (UTC)Reply
The difference is that it provides them a frictionless way to very quickly spam out edits they don't care about, and one that directs them to articles that already have problems, drowning out the people who actually do care about and are competent at fixing the problems. Gnomingstuff (talk) 22:13, 28 August 2026 (UTC)Reply
It's my experience that most newcomers actually do care about their edits. They may not be competent (yet), but most of us, including me, weren't competent in our early edits. WhatamIdoing (talk) 04:17, 29 August 2026 (UTC)Reply
@Elestrophe, this one looks like a WP:CIR case to me. I'm not sure they'd be any more or less annoying if Suggested Links didn't exist, honestly. If they don't change their tune in the next couple of days, feel free to ping me about it and I'll get them out of your hair. If there are individual math articles that are getting a disproportionate number of bad links, you can add {{No newcomer task}} to the article to keep them away. In solidarity, asilvering (talk) 21:36, 28 August 2026 (UTC)Reply
If Newcomer Tasks are based on maintenance templates then it's not surprising that they are poor because the maintenance templates are usually too vague and stale to be useful. In theory, they should be supported by talk page discussion which goes into detail but this is rarely done. And the actual talk page suggestions are often left dangling without being closed in a formal way.
There is a project called This week's article for improvement which is going to be featured on the main page soon. The idea is to encourage readers to become new editors. This will provide a good focus for improvement in the workflow for such tasks. I reckon that To do lists should be encouraged to provide a list of actionable tasks but I don't often see them on articles currently. The overall structure and workflow needs work.
I checked the latest one that caused me to start this discussion and the page did not have any template in place, nor do I believe the related ones that started to get on my nerves before that did either. They said in the replies that it is an LLM. ChompyTheGogoat (talk) 09:42, 29 August 2026 (UTC)Reply
Per Special:NewcomerTasksInfo, the vast majority of all newcomer tasks are link-recommendation. If I'm reading Special:CommunityConfiguration/GrowthSuggestedEdits correctly, I believe that is the "Add a link (Structured task)", which is not defined by templates - only excluded by them. revise-tone is also not mentioned there at all and therefore I assume that one is AI generated as well. For those tasks that are template defined, I'd recommend improving the suggestions by adding a parameter to define the level of work an article needs - maybe a 1-5 scale, with levels 4-5 automatically excluding them from newcomer tasks based on the level of work required, as Template:No newcomer task is meant to (which I've never seen utilized, and I would imagine most editors don't know it exists). Levels 1-3 could correspond to easy, medium, and hard tasks, instead of just assuming that all copyedit is easy. I would also recommend excluding articles with three or more maintenance templates for the same reason. ChompyTheGogoat (talk) 09:58, 29 August 2026 (UTC)Reply
18,000 articles are tagged with {{Promotional}}. How many of those are you personally willing to add a level-of-work parameter to?
have also been trying to explain this, as well as the fact that copy-editing skill and Wikipedia familiarity are not the same thing, and training someone to use the Wikipedia UI does nothing to improve their copyediting ability, especially if you are giving them unconditional virtual pats on the back via widget for doing such a good job Gnomingstuff (talk) 21:06, 29 August 2026 (UTC)Reply
The problem is that my experience has been that the newcomer tasks are not by any stretch of the imagination "an easy way to learn how to make an edit", but an easy way to waste hours looking into an issue (real or imagined) and ending up not making an edit or learning anything other than to avoid newcomer tasks. Yes there should be newcomer tasks. But they need to be very different from the ones I encountered. Long is the way (talk) 20:30, 26 August 2026 (UTC)Reply
They sound good in theory, but they reality is that they aren't good for helping people learn nor improving articles. They're confusing and waste editor time, especially when we have to clean up the "improvements that aren't". Like I said, they either need to be fully overhauled so they DO help, or removed to stop creating more problems. ChompyTheGogoat (talk) 20:57, 26 August 2026 (UTC)Reply
I think that most of them are helpful. But we don't have to wonder about which one of us is correct; we could set up the mw:ORES review tool and get some editors (all of us in this discussion?) to manually do a blind comparison of a random collection of newcomer tasks vs unprompted tasks by new editors. WhatamIdoing (talk) 16:12, 27 August 2026 (UTC)Reply
I'm not familiar with the tool. How exactly does it evaluate "overall quality"? I do believe that most newcomer task edits are good faith non-vandalism attempts to improve, but often not helpful because of the tool giving inappropriate suggestions and new users not having the experience to recognize that (or know what actually needs to be fixed). New editors who are vandals, UPE, etc wouldn't be likely to use the tool at all, so naturally more of those bad faith edits would be found without it. We'd need a narrower pool of test cases to avoid that bias. ChompyTheGogoat (talk) 21:26, 27 August 2026 (UTC)Reply
As a matter of fact, that could easily be skewing the existing metrics - purely the fact that most editors who'd use it are indeed good faith and WP:HERE. Maybe an examination of edits made with and without the tool by the editors who do utilize it? ChompyTheGogoat (talk) 21:28, 27 August 2026 (UTC)Reply
That tool works manually. It shows you a diff, and asks you what you think of it.
So imagine, e.g., that we set up this tool to show (without showing the Special:Tags) 10 edits from newcomer tasks and 10 similar-ish edits from equally inexperienced newbies that aren't from newcomer tasks. Then you rate them based on whether it's (in your best editorial judgement) a good edit or a bad one. WhatamIdoing (talk) 16:42, 28 August 2026 (UTC)Reply
Before ORES could produce those automated assessments (ORES is what color-codes watchlist items for "Likely have problems" and such), we had to feed it the original data. I guess the page I linked you to is more about the end result than about the tool for collecting the data, so that wasn't a very helpful link; sorry. WhatamIdoing (talk) 04:23, 29 August 2026 (UTC)Reply
So for a specific use like this editors determine which tasks are part of the pool to be evaluated, then it crunches the numbers for us - not just scanning and feeding us what it runs across in the wild based on given params? ChompyTheGogoat (talk) 04:35, 29 August 2026 (UTC)Reply
Well, more to the point, if we could resurrect the data-collection software that was used back then, we could feed it any set of diffs we wanted, editors could score them however they wanted, and we could crunch the numbers ourselves. WhatamIdoing (talk) 04:48, 29 August 2026 (UTC)Reply
Ok, so the current implementation of it doesn't allow us to designate a specific pool for evaluation? I thought that's what you were saying in your initial comment. ChompyTheGogoat (talk) 05:06, 29 August 2026 (UTC)Reply
I find it absolutely enraging that WMF would drop an AI feature onto English-WP as some sort of a beta test because somebody got a wack idea, a manager approved it, and engineers made work and developed it. I ran into a driveby "Newcomer Task: Suggested: Revise Tone" editor on a page I was actively working on that was flagged with a CONSTRUCTION template just yesterday. That is how I discovered the feature. That is how little regard that WMF paid staff has for the community and for the decentralized community decision-making process that has served us well for two decades. If it were up to the tech-worshiping, unforeseen-consequences-damning preferences of WMF, Wikipedia would by now approximate Grokipedia-With-Junkets. Something like this should NOT be unilaterally implemented by the engineers without community discussion. And I don't mean displaying notice of the forthcoming change in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying "Beware of the Leopard," either. Carrite (talk) 16:27, 28 August 2026 (UTC)Reply
The problem is that they were given a tool that encouraged them to spam out over 100 edits at a rate of roughly 1 every 3 minutes. No one fixed them for over two years, despite their very much needing fixing. Gnomingstuff (talk) 21:07, 29 August 2026 (UTC)Reply
The task itself is fairly simple: it highlights a paragraph containing language that the model has identified as commonly being reverted. In that respect, it is not all that different from the machine learning models that have been used for years to flag potentially problematic edits in Recent Changes. Revise Tone is also Community Configurable, so communities retain control over whether they want to offer the task. Any administrator can disable it if there is consensus.
Of course, I hope communities will choose to keep it enabled. Revise Tone, along with the other Newcomer Tasks, is intended to give people who are new to editing a relatively approachable way to make their first contributions. We know that getting started with Wikipedia editing can be difficult, and we need to provide newcomers with accessible ways to take that first step if we want to support the long-term sustainability of the projects. KStoller-WMF (talk) 22:35, 28 August 2026 (UTC)Reply
Community members were involved does not equate to consensus. This comes across (yet again) as "We're going to shove LLM down your throat, and if people protest loudly enough we'll consider removing it after the fact." And you wonder why editors are hostile to WMF involvement. ChompyTheGogoat (talk) 04:23, 29 August 2026 (UTC)Reply
I have asked for Newcomer Tasks to be disabled for several months now. I have done an audit on Newcomer Task quality -- the amount of good ones is dismally low. That's still true. I could do another audit, but I don't even know if that would help, because I have otherwise presented every piece of evidence I can possibly think of. There is no concrete evidence that anyone actually cares. (defined by anything actually being done about it beyond "we're listening") The situation is especially perverse for a number of reasons:
The articles hit by Newcomer Tasks are often articles that people tagged a long time ago. I assume that when they did so, their intent was not to make the articles worse, but that's what has happened.
The justification that we get, over and over, for why these are still around despite being a demonstrable net negative is the sunk-cost fallacy and how so much work has been put in. I don't know how to be any more polite here, but I don't care. If someone comes along and bashes a hole in my roof, I don't care how hard they worked to bash the hole, I care that my house is now being ruined by rain.
The other justification is that "well they're not being reverted so they must be good." The reason so many of them have not been reverted is because A) the firehose is spewing them out too quickly for "reverting" to even happen (in a way that puts the "reverted" tag on), and B) there are so many of them that everyone doing cleanup is swamped. The last time I brought this up I said I had over 100 tabs open with cleanup work. Now it's over 200. Just how fast am I expected to work to be able to make Number Go Down to a point that makes any impression whatsoever on the people who want see Number Go Up?
I don't remember anyone giving a sunk-cost justification.
Looking at Special:RecentChanges right now, I see just under 1,000 mainspace edits from newcomers (1–10 edits) in the last ~6 hours. About 11.5% of them are Newcomer tasks. 14% of them are already reverted. But: Only 3.7% of the Newcomer tasks are already reverted, whereas 16.2% of the non-Newcomer task edits have already been reverted. That's more than a fourfold difference. Newcomer task edits are only 23% as likely to get reverted as non-Newcomer task edits.
It might be that the daily ~4,000 mainspace edits from newcomers is more than our current Wikipedia:Recent changes patrol can handle. But it seems unlikely to me that the community is preferentially ignoring the Newcomer task edits, and if newbies using the Newcomer tasks are "only" as bad at editing as the rest of us were when we started, it would take a very significant level of ignoring edits to produce that big of a difference in the reversion rates. I therefore conclude that Newcomer task edits actually don't need to be reverted as often as other edits from newbies. WhatamIdoing (talk) 20:43, 28 August 2026 (UTC)Reply
Could you please respond to my numerous comments referring to the difference in editors who actually utilize the tool at all before you continue to rely on this logic? ChompyTheGogoat (talk) 04:32, 29 August 2026 (UTC)Reply
Sure: As has been pointed out repeatedly by multiple people in this discussion, Wikipedia:Long-term abuse socks, poop vandals, and other abusive actors might not look at Special:Homepage at all.
And as has also been pointed out, when you compare randomly assigned Group A, with the opportunity to complete newcomer tasks, against Group B, without that opportunity, and you see that Group A does the same or better than Group B on approximately every metric ever checked during the last ~seven years, even though most of the members in Group A never completed a task, then it's fair to assume that "actually using the tool at all" isn't necessary to get a benefit from it. The editors in Group A who use the tool are likely different from the editors in Group A who see the tasks and decide to edit independently (but who may have been influenced by the information they saw), and both of those subgroups are different from the editors in Group A who didn't look at the homepage at all, but there is no reason at all to assume that the randomly assigned members of Group A have a different number of abusive editors than the randomly assigned members of Group B, none of whom had the opportunity to see the page. And since Group A, including its voluntary non-users and its fair share of abusive editors, did much better than Group B, it's reasonable to think that the tool actually improves behavior overall. WhatamIdoing (talk) 05:02, 29 August 2026 (UTC)Reply
The Revise Tone experiment compared two different types of newcomer tasks, not a control group without any access to them at all. And the Add-a-link link is (ironically) broken (404). ChompyTheGogoat (talk) 10:05, 29 August 2026 (UTC)Reply
Yes, the individual subtasks have differing levels of value, which is why proposals, such as yours at the top of this thread, to remove all of them indiscriminately, are a bad idea. We should keep the ones we like, configure the ones we're okay with, and turn off the ones we dislike. WhatamIdoing (talk) 16:30, 29 August 2026 (UTC)Reply
@Gnomingstuff Thank you for taking the time to respond and for the previous audit work. I understand the frustration, especially the feeling that the cleanup burden has become overwhelming. Is there a particular task that you think is especially problematic, or do you feel that Newcomer Tasks as a whole are problematic?
I completely agree that we cannot equate “not reverted” with “good.” It is, however, one of the signals we have available. Looking at English Wikipedia article-namespace edits from January through July 2026, among editors with fewer than 30 days of tenure and fewer than 100 edits:
Newcomer edits overall had a 30.1% revert rate (900,292 of 2,994,419 edits). [1]
Edits made through Newcomer Tasks had a 4.7% revert rate (5,416 of 115,188 edits). [2]
There is an important caveat to this comparison: people who choose Newcomer Tasks are likely good-faith editors, while the overall newcomer figure includes vandalism and other clearly problematic newcomers. So this comparison almost certainly overstates the difference. And, as you point out, neither number captures cleanup that happens without a formal revert. Even with those caveats, though, the data suggests that Newcomer Tasks are not disproportionately contributing to the revert workload relative to newcomer editing overall.
The reality is that newcomers will always make some mistakes as they learn. That is part of bringing new people into the project. At the same time, I do think there are several promising projects underway that should help:
Better onboarding and task matching: The Growth team is working on early onboarding changes, including changes to the Newcomer Task feed. We are working toward a smaller, more curated set of tasks that better matches contributors to tasks based on skill level and interests.
More accurate Add a Link suggestions: The Machine Learning team is working to improve the accuracy of Add a Link suggestions, which should reduce the number of poor-quality suggestions reaching newcomers (T434259). The Growth team will also work on an Add a Link improvement soon that will both decrease the quantity of suggestions available and also increase the quality of suggestions (T429417).
More learning support: The Community Development team is working on "micro-learning" videos based on the Wikimedia Core Curriculum, giving newcomers more guidance at the point when they need it.
Catching mistakes before publication: The Editing team's Edit Check work catches some common mistakes before they are published.
Expanding the moderator pool: The Moderator Tools team is working on ways to onboard moderators and patrollers, so that the work of reviewing newcomer edits can be distributed among more people.
None of this clears your 200 tabs today, and I don't want to pretend that it does. But the direction we're investing in is fewer, better-matched tasks, with more support for newcomers and better safeguards around the edits they make, while also helping newer editors develop toward appropriate moderation and patrolling roles.
@KStoller-WMF, would you also consider coming up with some better way to categorize articles for the purposes of showing "relevant to your interests" articles to newbies? There's someone complaining about that upthread, and I recall also finding this pretty useless for the same reason. I was under the impression that those ORES topics were going to be replaced by something much better years ago, and that hasn't happened. Is anyone still working on that? In solidarity, asilvering (talk) 21:42, 28 August 2026 (UTC)Reply
Design showing how newcomers could select articles of interest before arriving to their Homepage
We hope to release an A/B test as early as next month in which we actually allow newcomers to select articles of interest to populate a more limited set of suggestions. Related project page: https://www.mediawiki.org/wiki/Home
@KStoller-WMF, this is much better!! All the articles it gave me are related to my interests, and none are in quite such a horrible state that it's depressing to look at them. And, well, it couldn't have known, but... it suggested one of my own articles (Richard Caudray) for expansion. In solidarity, asilvering (talk) 22:26, 28 August 2026 (UTC)Reply
My suggestions are also much better, but I can't click through to check on anything. The crosslinks and references tasks seem reasonable - I'm not sure what "bring up to date" is looking for, which sounds rather vague. I didn't get any revise tone suggestions, which I honestly feel is better because newbies don't have a good feel for Wikipedia voice and NPOV yet - especially since it's marked an "easy" task, which to me should be the very first edits someone ever makes. Crosslinks and basic copyedit are about as easy as it gets. (Crosslink suggestions aren't always accurate or needed, but very low in terms of the actual problem they cause.) ChompyTheGogoat (talk) 05:01, 29 August 2026 (UTC)Reply
I'm going to WP:AGF and assume that a WMF employee knows better than to use LLMs in discussions - we know these tools aren't fully accurate - but I cannot stress enough that if you're spending so much time engaging with AI that you start to sound like them you should seriously check yourself. ChompyTheGogoat (talk) 04:48, 29 August 2026 (UTC)Reply
All of the tasks are net negatives in practice with the exception of Suggested Links, since the damage an individual editor can do with it is very small and contained, and does not affect the prose.
The type of the task also doesn't matter, as people disregard it all the time. Just a few examples taken from the hundreds of tabs I am slogging through:
Special:Contributions/HelloHop (Note the **Suggested edit-summary (copy-paste into the “Edit summary” box):** chatbot response in one edit summary, an indication of how thoroughly this person reviewed the edits they have spammed out)
Let's not forget that their response to sunk cost concerns is to barge ahead anyway instead of pumping the brakes, so I have zero sympathy for that at this point. If you want to ensure your work will have a lasting beneficial impact, make sure it's something anyone bloody well wants before you ram implementation through. Or eat the consequences. ChompyTheGogoat (talk) 04:30, 29 August 2026 (UTC)Reply
Sunk cost fallacy is one of those power words that power users like to throw around, usually when they have a gut-level revulsion to a change and don't think they'll be able to stop the change. There's a relevant source linked at the end of Wikipedia:You don't own Wikipedia that might prove to be interesting reading, if you haven't seen it before.
But, as a point of fact, the stated response in that discussion (which comes from a long-time Wikipedia editor, BTW) is that they've designed this project so they can easily "abandon" anything that doesn't look promising. A lot of the ideas the WMF evaluates don't see the light of day (e.g., the most recent round of "let's change the font!", which comes up every five years or so – but never from someone who lived through the last attempt), so you probably wouldn't hear about them unless you watch phab: regularly. Consequently, it is important not to assume that the continuation rate for the few projects you've heard of is the overall continuation rate. WhatamIdoing (talk) 05:23, 29 August 2026 (UTC)Reply
Perhaps not, but it seems to be their preferred response on these related subjects. It was clearly communicated that numerous editors are uncomfortable with that project and a desire for consensus was expressed, and the response was "we can always abandon it later", which does nothing to address the underlying concern. Do they believe the community is going to magically change our opinion on LLMs by that point, or are they then going to argue that they should keep going after putting so much work into it? What harm does it do to pause and see whether people really do want this tool at all, instead of "what suggestions can be used going forward because we're definitely going forward"? It feels like they're willing to reconsider specific aspects based on input, but not the project as a whole. Rather WP:IDIDN'THEARTHAT of them, IMHO. ChompyTheGogoat (talk) 05:57, 29 August 2026 (UTC)Reply
No, but they might believe that (a) people who haven't tried the tool don't have the information they need to make an informed decision, and (b) that even if the tool is abandoned, something useful could be learned from it. And, of course, we all know that (c) the ~five dozen people in that discussion, some of whom support the project, are not even remotely representative of the three-quarter million registered editors who make at least one edit in a given year.
More generally, we have the problem that (d), if you reach out to the communities early in a project, when it would be cheap and easy to abandon it, then people don't understand the project or its goals, and even complain that you brought the idea to them so early, when you don't even know how it will behave or whether it works or what it looks like. But if you bring it to them later, when you can provide solid answers to most of their questions, they say "How dare you not involve me early in the process! Nobody asked for this! (Pay no attention to those diffs behind the curtain that prove that someone else in the community did ask for it.) It wasn't discussed! (a common enough complaint that experienced people create lists of prior discussions such as Wikipedia:Vector 2022#List of discussions, but you still get nonsense, like this IP claiming "that nobody saw" an RFC that 344 people participated in) I hate it! (but a couple of months from now, I'll probably have gotten used to it). It was before your time, but the RFCs for Vector 2022 was so predictable (and predicted) that I should have started a betting pool on when the first rollback RFC would be started, measured in hours after deployment.
The WMF product folks who are involved in the project being discussed likely remember my views on anything that sounds like Microsoft's Clippy: I'm not a fan. But I don't think that the work they're doing is useless, or that any editors will be forced to use it. WhatamIdoing (talk) 06:42, 29 August 2026 (UTC)Reply
Consensus is never expected to require input from the entire editor pool, just enough who have interest in the specific proposal to get a solid feel for the overall sentiment based on rational arguments (especially, but not exclusively, made by senior editors who do understand both the relevant guidelines and history of the subject on-wiki). And I don't think early consensus should be the final say on wide scale implementation, but an indication that most people think the general concept has enough promise to be worth developing and evaluating once it's functional. If they built a full Grokipedia style bot to write unreviewed articles and didn't listen to the community until it was in on-wiki testing I expect there would be a few opinions. I also did read your previous reference to Wikipedia:You don't own Wikipedia and I really don't feel it applies in this situation. Obviously not to me - I haven't even been here for a year, but I've seen how much harm AI can do both on wiki and off, and I've seen how the general community feels about its implementation, as reflected in existing guidelines. I've also seen similar sentiments on a smaller scale when it comes to newcomer tasks, which is why I started this, and I'm not at all surprised to learn it's also AI because the randomness and low quality of the suggestions is exactly what I expect from slop that can't comprehend the task at hand because it has no comprehension. An RfC would garner input from a wider variety of editors so no "power users" can attempt to strongarm their opinion through. Obviously the WMF should and does have ultimate authority when it comes to legal issues, finances, etc - but that's not what this is. Individual projects are supposed to have a wide degree of latitude for how to manage content that doesn't cross any legal lines, and English wiki is largely against AI implementation. If they want to develop it for different projects that have wider acceptance, have at (but I suspect they wouldn't devote the resources to it if we reject it from EN). ChompyTheGogoat (talk) 09:38, 29 August 2026 (UTC)Reply
@ChompyTheGogoat, the way WAID's essay applies here is that we desperately need more new editors to step up, to replace those of us who inevitably will wander away from the projects (or die in office). It's going to be a group effort to make the projects more welcoming to newcomers, and this is one possible way. AI has some real promise in surfacing appropriate tasks for newcomers, because of the WP:SOFIXIT attitude that longtime editors have - if we spot something easy, we just fix it ourselves. That means that what's left - the stuff that's tagged - is often an absolutely horrible slog to deal with. That was my own experience of the newcomer homepage, when I started - it was entirely the tasks generated by maintenance templates, and what it sent me to was articles so broken that I felt completely demotivated to try to fix them. The tool was suggesting I get some basic editing experience in, and sending me to articles that needed to be completely rewritten, not things that were a quick job at all. Things like the suggested links task and revise tone can help find spots that need help that are much more within the capabilities and inclinations of newbies. In solidarity, asilvering (talk) 16:58, 29 August 2026 (UTC)Reply
The problem with a non-representative group of editors is that you don't "get a solid feel for the overall sentiment"; you instead "get a solid feel for the overall sentiment within a non-representative group, which may or may not differ significantly from the overall sentiment of the whole community". Sometimes there's no important differences; that's why most RFCs, with a typical participation of 5 to 15 editors, work. But sometimes it does matter.
I suspect that the bigger weakness with the Small language model is that it can't compare article content against source content, so calling William Shakespeare "the greatest writer in the English language and the world's pre-eminent dramatist" will seem puffy, and that saying Martin Shkreli has a "reputation as 'the most hated man in America'" will seem disparaging. But both of these are justified by the sources, and the SLM machine learning tool has no way of knowing that. WhatamIdoing (talk) 17:16, 29 August 2026 (UTC)Reply
I'm not at all surprised to learn it's also AI
Just to clarify, the problem is that people are using AI to spam out the tasks. If people weren't using AI to spam out the tasks, there would be less of a problem. But the tool itself encourages this behavior, by design:
The gamification system, by design provides an incentive for them to do so as fast as possible so Number Go Up as fast as possible, and provides repeated praise, implicitly and explicitly, as they do it.
The tool funnels them to articles that have already been identified as problematic, making those articles worse: a slap in the face to everyone who tagged articles for improvement in good faith because they wanted them improved.
The tool funnels multiple editors at a high speed, meaning that those articles' edit histories become so drowned beneath bad edits that none of them can be reverted without painstaking work. Essentially, it gives less-trafficked articles the edit volume of something on the front page, except without the people watching it.
I have the impression that Chompy's concern is that the "Revise tone" task is using a Small language model to find pages for its suggestion list.
I've just done six Revise Tone tasks. Five needed help, sometimes badly. The other was correct. It only proposed changes to a single paragraph at a time. It promptly asked me to switch to a more advanced task. If we're concerned about people doing too many of these in a single day ("as fast as possible so Number Go Up as fast as possible"), then maybe we should ask for daily limits. WhatamIdoing (talk) 00:32, 30 August 2026 (UTC)Reply
Drive-by comment:
I skimmed through the overly tedious conversation. Apparently, the filing editor wants to upgrade the newcomer task feature or remove it, citing concerns over whether the tool is accurate, and how the new editors leave out work for the more experienced editors to fix. While I support a better system, it appears that this is just a WP:Competence is required case. Also, reading through, I am unsure what exactly they want to change. Further, this newcomer system is flawed by design, making a negative feedback cycle.
Is it possible that some PendingChanges type of monitor list shows a subset these edits for human review? Or, change the system so there is a complete Wikipedia course, something that is quite lacking to newbies. 16dvnk (talk) 11:46, 1 September 2026 (UTC)Reply
BTW just a heads up, the WP:JCW compilation now reports whether or not publications and publishers are open access ones. It's not perfect, sometimes there are classes like (delayed open access being the category given to the article, but the infobox just says open-access = yes, so the bot picks one which might not be accurate. Most publishers should be unflagged because the publisher a mix of open and closed access journals, while most publications should probably be flagged as hybrid open access, since most will support a pay-to-make-it-free option.
