Wikipedia talk:Template index/User talk namespace

Uw-ai alternative for uw-ai-image

Hello, is there any interest in rephrasing uw-ai to also reference AI generated images? I would assume we want to have a quick way to tell users that AI generated images do not add value or may have an underlying copyright that makes them not reasonable on Wikipedia. Not sure all the policies yet on this beyond no LLM use in editing. --Engineerchange (talk) 18:00, 10 June 2026 (UTC)

I use {{uw-image1}} and add a short note mentioning WP:AIGI or WP:AIUPSCALE as appropriate. A {{uw-aiimage}} template that already included that for me would certainly be useful. Belbury (talk) 18:11, 10 June 2026 (UTC)

@Engineerchange and Belbury: There are a couple different ways you could go with this. One is to have a whole, new template, or to modify either {{uw-ai}} to mention images, or {{uw-image1}} to mention AI. I've mocked up a quick sandbox version to give you an idea what it might look like:

Illustration of a modified version of {{uw-image1}}     (two examples)

These two examples are from the wording in sandbox rev 1359623276.

1. This example was substed, but you can test it in sandboxes transcluded or substituted.

{{subst:Uw-image1/sandbox|ai=yes}} ⟶

Hello! Your image was inserted successfully but because it appeared to be irrelevant to the article or violated the image use policy on AI-generated images, it has been reverted or removed. Please use the sandbox for any tests you want to do. Take a look at the welcome page if you would like to learn more about contributing to our encyclopedia. Thank you.

---

2. substed example with param1 present:

{{subst:Uw-image1/sandbox|Underwater basket-weaving|ai=yes}} ⟶

Hello! Your image was inserted successfully on the page Underwater basket-weaving, but because it appeared to be irrelevant to the article or violated the image use policy on AI-generated images, it has been reverted or removed. Please use the sandbox for any tests you want to do. Take a look at the welcome page if you would like to learn more about contributing to our encyclopedia. Thank you.

More illustrations of {{uw-image1/sandbox}}     (three examples from sandbox rev. 1359878988)

New wording as suggested below @09:10, 17 June et seq.; from sandbox rev 1359878988.

1. Default wording (no params); note: the wording in sentence one is re-ordered from live: the action ('reverted or removed') now appears before the reason ('because it appeared to be...' )

{{subst:Uw-image1/sandbox}} ⟶

Hello! Your image was inserted successfully but it has been reverted or removed because it appeared to be irrelevant to the article or violated the image use policy. Please use the sandbox for any tests you want to do. Take a look at the welcome page if you would like to learn more about contributing to our encyclopedia. Thank you.

---

2. With param 1 (article name) only (no ai param):

{{subst:Uw-image1/sandbox|Underwater basket-weaving}} ⟶

Hello! Your image was inserted successfully on the page Underwater basket-weaving, but it has been reverted or removed because it appeared to be irrelevant to the article or violated the image use policy. Please use the sandbox for any tests you want to do. Take a look at the welcome page if you would like to learn more about contributing to our encyclopedia. Thank you.

---

3. With param 1 (article name) and param ai=yes:

{{subst:Uw-image1/sandbox|Underwater basket-weaving|ai=yes}} ⟶

Hello! Your image was inserted successfully on the page Underwater basket-weaving, but it has been reverted or removed because it appeared to violate the image use policy. Except in AI-related topics, articles should generally avoid AI-generated or substantially AI-altered images, especially of named individuals or scientific subjects. Where AI alteration has occurred, the original image is preferred. Please use the sandbox for any tests you want to do. Take a look at the welcome page if you would like to learn more about contributing to our encyclopedia. Thank you.

The language could easily be altered to whatever you want for the |ai=yes case, so feel free to make suggestions or just modify Template:Uw-image1/sandbox directly. Thoughts? Mathglot (talk) 08:19, 16 June 2026 (UTC)

@Mathglot: Your example with the |ai=yes param meets all of my needs. Thank you, --Engineerchange (talk) 10:48, 16 June 2026 (UTC)

Engineerchange, I'm glad. But in theory at least, one shouldn't use the sandbox version yet at a User talk page, as it could change at a moment's notice, and also is not a live version with consensus. We should wait a bit to see if anyone else chimes in to offer alternate wording, or to object to the change. If there's nothing in say, a week, then we can move it to live. Would you mind tracking this and then pinging me in a week if it looks like we can release it? Thanks, Mathglot (talk) 19:46, 16 June 2026 (UTC) P.S. I added another example above showing use with param1=article name. Mathglot (talk) 20:07, 16 June 2026 (UTC)

@Mathglot: Can do! Thanks, --Engineerchange (talk) 23:35, 16 June 2026 (UTC)

@Mathglot: Been a week. Can push for another week if you think there's still dialogue here. --Engineerchange (talk) 13:06, 23 June 2026 (UTC)

Engineerchange, yes, let's. (Thanks for the reminder.) I did sandbox versions for them, but will need to redo two of them (4 and 4im) partly for technical reasons as they are structured differently internally. Also, I am creating test cases for all of them so you and others can view them side-by-side with the originals; it would be too much to do all that in this discussion. You can look at 2 and 3 now, at Template:Uw-image2/testcases and Template:Uw-image3/testcases (in particular, test 5a and 5b), but there's no explanation of the tests or any new template doc yet. Templates 4 and 4im need to be redone, with new test cases. The main thing to note is that param 2 overrides the new AI param, so if you have |2=, then |ai=yes is ignored (because if you are writing your own text, then you are writing your own text; can't have it both ways). Mathglot (talk) 00:55, 24 June 2026 (UTC)

Mathglot About another week, so sending a reminder. Happy to test when you find the time. I'm in no rush, just want to keep the thread alive. And what you say about the 2nd param not working makes sense to me? I feel like as long as it doesn't override param 1, it's not a concern in my view. --Engineerchange (talk) 04:08, 2 July 2026 (UTC)

I think we can drop because it appeared to be irrelevant to the article or if we know that the template is being added because of the AI issue. It may also save some time and questions if we gave the user a short summary of WP:AIGI in the text of the template, something like Unless the article is about an AI-related subject, Wikipedia articles should generally not use images generated or substantially altered by AI, especially when depicting named individuals or scientific subjects. Where an image has been altered by AI, the original version is preferred. Belbury (talk) 09:10, 17 June 2026 (UTC)

That seems like a good goal. Can we tighten it up/shorten it a bit thus:

Except in AI-related topics, articles should generally avoid AI-generated or substantially AI-altered images, especially of named individuals or scientific subjects. Where AI alteration has occurred, the original image is preferred.

I don't think that loses anything essential in your version; what do you think? Mathglot (talk) 19:49, 17 June 2026 (UTC)

Sure, good one. Belbury (talk) 19:52, 17 June 2026 (UTC)

Please see "More illustrations..." above, in the dusky green collapse bar, with new wording per recent suggestions. There is also an additional mod to sentence 1 for more natural flow of action because reason, unrelated to the ai wording discussion. Mathglot (talk) 21:06, 17 June 2026 (UTC)

Looks good.

Having just had a user question this sentence in uw-image, on their talk page, Please use the sandbox for any tests you want to do. actually doesn't seem relevant when talking about AI images, so maybe it's worth losing it here. Belbury (talk) 16:23, 18 June 2026 (UTC)

You mean, drop it only in the ai=yes case, right, or in both versions? Mathglot (talk) 19:18, 18 June 2026 (UTC)

• Listed at: Wikipedia talk:Artificial intelligence and Wikipedia talk:Image use policy. Mathglot (talk) 19:57, 16 June 2026 (UTC)

Engineerchange, one other thing: if this does get approved, then it raises the question what, if anything, to do about {{Uw-image2}}, {{Uw-image3}}, {{Uw-image4}}, and {{Uw-image4im}}. I had a look at them, and assuming the users receives the level-1 warning first with |ai=yes, I'm not sure it needs to be repeated again for levels 2 and higher. Have a look, and see what you think. If you think we do need ai=yes versions for those as well, it's really n.b.d., but in that case, please suggest what the wording should be for each one. Thanks, Mathglot (talk) 08:11, 17 June 2026 (UTC)

@Mathglot: My suggestions (all wikilinks remain the same as normal):

• {{Uw-image2}} Please do not add inappropriate images to Wikipedia; it is considered vandalism. Please see the image use policy for appropriate use of AI-generated images in Wikipedia. Thank you.
• {{Uw-image3}} Please stop. If you continue to add inappropriate images, you may be blocked from editing. Unencyclopedic images, including AI-generated images, may be added to the bad image list to prevent them from being displayed.
• {{Uw-image4}} You may be blocked from editing without further warning the next time you add an inappropriate image. Unencyclopedic images, including AI-generated images, may be added to the bad image list to prevent their display if used persistently for vandalism.
• {{Uw-image4im}} This is your only warning; if you add an inappropriate image to Wikipedia again, you may be blocked from editing without further notice. Unencyclopedic images, including AI-generated images, may be added to the bad image list to prevent their display if used persistently for vandalism.

Namely just adding a simple note that AI-generated images are what is unencyclopedic for 3, 4, and 4im. Also, added an additional clarifying policy point to 2. IP and new users need this type of repetition in these warning messages, in my experience. But very open to edits here. --Engineerchange (talk) 14:32, 18 June 2026 (UTC)

Revised the above given the sandbox statement further above. --Engineerchange (talk) 16:29, 18 June 2026 (UTC)

@Engineerchange and Belbury:, sorry, I've been busy, but I think we may have it, now. The changes are too numerous to include tests here, so you can either test the sandbox pages at {{Uw-image2/sandbox}}, {{Uw-image3/sandbox}}, {{Uw-image4/sandbox}}, and {{Uw-image4im/sandbox}} in your sandbox or at Special:ExpandTemplates, or look at the test case pages for Uw-image2, Uw-image3, Uw-image4, and Uw-image4im. This wasn't conceptually difficult, there's just a lot of little details to keep track of all over, and it would be easy for me to have missed something somewhere. So please be an extra set of eyes, and make sure everything looks good, and if you spot anything fishy, please lmk. If it all looks good, then if we wait a few more days for any additional comment, we can release them. Mathglot (talk) 02:23, 22 July 2026 (UTC)

Oops, sorry; found a systematic problem to fix. Don't bother testing yet; I'll let you know. Mathglot (talk) 02:28, 22 July 2026 (UTC)

Actually, the problems only concern 3, 4, and 4im which transclude {{Uw3}}, {{Uw4}}, and {{Uw4im}}, whereas 1 and 2 emit the messages directly, and they are both fine, afaict. If you don't see a downside to releasing the new AI messaging-capable versions in two batches (i.e., a: 1 and 2; and b: 3, 4, and 4im), then we can do 1 and 2 now, assuming you are good with the outputs. In that case, please see the test case pages for Uw-image1/testcases and Uw-image2/testcases, and do whatever additional checks you like, and if you are on board, I will release those two. (Tests #5a – 5d are the ones to watch; the sandbox versions produce the new messaging; the live versions produce the old messaging.) Those two templates constitute the overwhelming majority of all warnings from the series anyway: there are 213 placements of 1 and 2 so far in 2026, and only 63 so far for 3, 4, and 4im combined. Thoughts? Adding @Engineerchange and Belbury:. Mathglot (talk) 21:34, 22 July 2026 (UTC)

Warning against using AI to appeal a block

Probably half the unblock requests I review have been obviously generated with AI/LLM. Although it doesn't take long to decline such a request with the canned message available, and in that sense this is perhaps 'no biggie', it just seems so utterly pointless and a complete waste of the user's and the admin's time. I don't know if the users – esp. those likely to resort to AI – actually even read the block notice, but assuming they do, could we add to the template a warning against using AI? -- DoubleGrazing (talk) 08:00, 23 June 2026 (UTC)

As with other prohibitions of this type on Wikipedia, I wonder if it is really against using AI, or it is against using obvious and low quality AI. We could perhaps say "Low effort, vague, or incomprehensible unblock requests, including those generated by a 'chat bot' or LLM, may be summarily declined, or be considered to aggravate the reasons for a block." —DIYeditor (talk) 08:25, 23 June 2026 (UTC)

I for one (and YMMV, of course) am categorically against using AI at all in this context. When the user copypastes from a chatbot something like "I fully understand X, agree to do Y, and promise not to do Z", there is no way of knowing whether they even read what they pasted into the appeal template, let alone actually understood it. And given that the point of the appeal process is to gauge whether it is safe to unblock the user with the expectation that they will not carry on the behaviour which led to the block, this seems to me to defeat the whole purpose of the exercise. -- DoubleGrazing (talk) 08:52, 23 June 2026 (UTC)

There's always an implicit "if we can tell" in these prohibitions. If the AI is good enough, you are actually free to use it, because nobody could tell the difference. So the real problem is low effort or nonsense material. —DIYeditor (talk) 19:03, 23 June 2026 (UTC)

A key feature of an appeal is understanding why you were blocked. Using AI to formulate an appeal either transfers this understanding to the bot (if you believe that bots can 'understand' something) or implies that nobody understood anything. Either way, this renders the appeal null and void, and it should be rejected immediately without consideration.

A secondary question might be posed:

I am an ESL speaker, and I wrote the appeal myself in my sketchy English, and I only used AI to fix up the grammar, spelling, and style.

My response here would be:

Non-native English speakers are welcome here. You are welcome to use your own, imperfect English in a block appeal; please do not use AI to fix the grammar or style. Either your English is comprehensible, in which case there will be no prejudice to your appeal, or else it is at a level where a reasonable person just can't be sure they understand what you mean. In the latter case, you should not be editing here at en-wiki; but did you know that there are Wikipedias in 300 languages? You are welcome to edit there.

Thanks, Mathglot (talk) 21:56, 22 July 2026 (UTC)

I have zero objections, but I feel a need to remind everybody we live in a world where it is absolutely not obvious to people why we are so dead set against AI. By which I mean we should not take it for granted people understand please do not use AI to fix the grammar or style, I could certainly see someone using AI out of good faith and not maliciously; maybe even better than "please do not use AI to fix the grammar or style" would be "please do not use AI to fix the grammar or style (if you wonder why we ask you to not use AI: LINK HERE)" CapnZapp (talk) 12:08, 23 July 2026 (UTC)

Legal action block template

I guess I just noticed that the wording had been changed recently, removing "or the legal action is unresolved.". I understand that a legal action between two editors would not prevent them from editing as long as no threat was made on Wikipedia(even though any lawyer would tell their clients not to engage with the person they are undertaking a legal action against, but yeah, that's not our concern). I have two questions:

1. If the legal action is against Wikipedia or the WMF itself, shouldn't the person remain blocked until the action is concluded?
2. If someone makes a legal threat, and is blocked, does them carrying out their threat mean they can be unblocked? (it's no longer a threat if they carry it out) That seems like a perverse way to encourage legal action.

331dot (talk) 12:10, 6 July 2026 (UTC)

Pinging @Oshwah as the editor who made the change. I think both wordings can be interpreted to say that a legal threat is a legal threat even while there's subsequent legal action, until that action is resolved or the threat is revoked. The old version could be read to mean you were blocked only for taking legal action against another user, even if you hadn't made any on-wiki legal threat. The spirit of the rule is that the "I'm gonna sue you" card is so disruptively uncivil as to warrant a block. The disruption is what's being blocked, and the disruption continues while they are following through on the threat. I don't see any reason for threat-turned-actions against Wikipedia or WMF to be treated differently.

As for whether someone should be blocked for taking legal action against WMF without making an on-wiki threat to do so, if the Office wants to block someone they're being sued by they will take that action themselves. tony 17:39, 6 July 2026 (UTC)

331dot, TonySt - TonySt said it correctly. I was under the belief that legal threats were not permitted on Wikipedia. As NLT states: "The existence of a legal dispute between users, whether as a result of incidents on Wikipedia or elsewhere, is not a valid reason to block, as long as no legal threats are made on Wikipedia." (Emphasis mine) This is why I changed the wording, as unresolved legal action without a threat being made on Wikipedia is not a valid reason to block. ~Oshwah~^{(talk) (contribs)} 16:40, 8 July 2026 (UTC)

Thank you both very much. :) 331dot (talk) 16:49, 8 July 2026 (UTC)

331dot - You bet! It's good to talk to you again; I hope all is going well with you. ;-) ~Oshwah~^{(talk) (contribs)} 03:50, 9 July 2026 (UTC)

Template-protected edit request on 17 July 2026

This change would make the bold talk page text in the template a wikilink to the talk page of the article passed as a parameter. If the article parameter is unfilled, it keeps the current functionality. Currently, the text links to talk page etiquette guidelines, but the next sentence still does that too. My changes can be found here: Template:Uw-3rr/sandbox.

−

Instead, use the '''[[Wikipedia:Talk page guidelines|talk page]]''' to discuss how the page can represent [[Wikipedia:Consensus|consensus]] among the editors involved.

+

Instead, use the '''{{<includeonly>safesubst:</includeonly>#if:{{{1|}}}|[[{{TALKPAGENAME:{{{1}}}}}|talk page]]|[[Wikipedia:Talk page guidelines|talk page]]}}''' to discuss how the page can represent [[Wikipedia:Consensus|consensus]] among the editors involved.

FlammablePizza (talk) 15:45, 17 July 2026 (UTC)

 Done. Sounds good to me. Axolitl (talk | contribs) 03:35, 3 August 2026 (UTC)

Edit request 2026-07-24

The extraneous space between "any contributions made while evading blocks or bans" and "may be reverted or deleted" should be removed. Ascchrvalstr (talk) 13:57, 24 July 2026 (UTC)

 Done in Special:Diff/1366135317. —⁠andrybak (talk) 13:43, 26 July 2026 (UTC)

New warning template {{Uw-pageswap}}

I run into this issue with some regularity, so I created a single notice. I'd like someone to check my work, as it were. It has two required parameters so the boilerplate documentation does not quite fit it. It also has an optional parameter adding a sentence on broadcasting moves, which are a common venue for attempted pageswaps. Sammi Brie (she/her · t · c) 19:20, 31 July 2026 (UTC)

Hello, Sammi. That seems like a worthwhile issue to templatize, thanks. Took a brief look, and no screaming red flags jumped out at me but before looking at it more seriously, you need to create a testcases page. I always use template {{test case}} for that, which is powerful and has lots of display options. Not everybody does, though; different editors have somewhat different styles for creating test case pages. You can see some examples of mine at {{Uw-draftnocat/testcases}}, {{Uw-image2/testcases}}, {{Wrapper/testcases}}, (or {{WikiProject status/testcases}}—hybrid; partly mine), but check out other styles at any of these uw-* testcase pages. (Note: the testcase template requires a template sandbox to compare to; for now, you can create the sandbox as an exact duplicate of the template.) Once you have your testcases page, write again. Mathglot (talk) 22:49, 4 August 2026 (UTC)

@Mathglot Created testcases. This template is different from most others, as it takes two pagename parameters instead of one (documentation doesn't line up). You can't just swap one page, after all. Sammi Brie (she/her · t · c) 00:10, 5 August 2026 (UTC)

Sammi, but not different from all others; see e.g., {{Uw-articletodraft}}, {{Uw-draftmoved}}, {{Uw-translation}} for how they do it. (P.S., when you are discussing a page, please make it easier for others by linking the page; e.g. {{Uw-pageswap/testcases}}.) Thanks for starting the test page. Your test 2 works in theory, but you wouldn't likely be swapping such disparate names as Underwater basket-weaving and Macramé, so while the test is valid, it might be pretty confusing to viewers; also, the test name could be better, as both params 1 & 2 are article names, right? Test 5 tests appear to be copypasta which apply to parameters not in your template; e.g., you have no param 'broadcast', right? Take another shot. I may be busy for some days, but you're making good progress; ask for feedback at projects or noticeboards. Mathglot (talk) 00:29, 5 August 2026 (UTC)

Edit request for Template:Uw-biog1

Currently, Template:Uw-biog1 appears to contain two grammatical errors. First, there's an erroneous comma—separating the verb phrases in a compound predicate—after the first instance of the word "person."

I noticed that you changed content about a living or recently deceased person, but didn't support your changes with a reliable source.

Second, the sentence starting with "Wikipedia has strict policy on" is missing the article "a" before the noun "strict policy."

Wikipedia has strict policy on how we write about living people to keep such articles accurate and clear [...]

I propose the following change to fix both of these errors.

−

I noticed that you changed content about a living or recently deceased person, but didn't support your changes with a reliable source. Wikipedia has strict policy on how we write about living people to keep such articles accurate and clear, so your change has been removed and archived in the page history for now.

+

I noticed that you changed content about a living or recently deceased person, but you didn't support your changes with a reliable source. Wikipedia has strict policies on how we write about living people to keep such articles accurate and clear, so your change has been removed and archived in the page history for now.

 Done These are great changes. As someone who has WP:CINS in my userspace, I heartily support them. Sammi Brie (she/her · t · c) 00:10, 5 August 2026 (UTC)

New User Warning Template

 Courtesy link: Template:Uw-aiask

I made a new user warning template and was wondering if it would be okay to add it to the list. When editing, especially with COI edit requests, I frequently encounter situations where a editor appears to use AI but I can't definitevely say. I made this template to ask people whether or not their edits were AI generated. Thanks! - Otherwise (Antibacklog Aktion!) (talk?) 15:08, 4 August 2026 (UTC)

Mustbeotherwise, You can't really tell someone (in bold, no less) to refrain from editing further until you answer this message; that's pretty much a no-go. Secondly, are you aware of some existing messages that deal with AI issues, and how do you see your new one fitting in to existing ones?

Secondly, you might ask for more feedback about this by posting a very brief message at the WP:AI noticeboard and linking this discussion and asking for opinions here. P.S., I added a courtesy link at the top, because your message did not name it in plain text. Mathglot (talk) 22:58, 4 August 2026 (UTC)

Youre right that I should've added a courtesy link, my bad.

Ill add a link on the notice board when I get to my PC (currently on mobile)

I took a cue from Template:Uw-paid1 which says In either case, do not edit further until you answer this message. Is the distinguishing feature that the AI related rules are guidelines and the Paid editing rules are a policy? I didn't realize that it would be a point of contention. Thank you for your feedback! - Otherwise (Antibacklog Aktion!) (talk?) 23:03, 4 August 2026 (UTC)

also, to answer one of your points that I didn't address in my prior message, all AI talk page templates currently in the table of templates presume that you have used AI. I think it would be very helpful to have a template that asks if you have used AI or not in cases where you're not sure, or in cases where a more diplomatic approach is warranted. - Otherwise (Antibacklog Aktion!) (talk?) 23:06, 4 August 2026 (UTC)

@Mustbeotherwise As someone who is seeing the same AI tidal wave crashing down on COI edit requests, I understand this. You don't want to BITE people who are unfamiliar with the project. The "refrain from editing until you do this" makes sense for paid editors because of the rules we have there. It doesn't make sense here. A lot of people that might be used to using AI for work these days, or using AI to cope with the byzantine world of Wikipedia, are coming here and not realizing that AI use is frowned on. Sammi Brie (she/her · t · c) 00:13, 5 August 2026 (UTC)

Yes, the {{uw-paid}} template does say that, and I doubt it should. At most, it should be limited to articles related to their COI, imho, but I'm just one person, so it's good to get a variety of opinions.

With regard to this:

I think it would be very helpful to have a template that asks if you have used AI or not in cases where you're not sure, or in cases where a more diplomatic approach is warranted.

there is a similar dichotomy between {{uw-coi}}, and {{uw-coi-brief}} (which should probably be named {{uw-coi-ask}}). Mathglot (talk) 00:37, 5 August 2026 (UTC)

Dear @Mathglot and @Sammi Brie,

I appreciate both of y'all's feedback. At the time I created the template, I did not realize that {{uw-paid}} was relatively unique in its assertion to not edit further. I plan on rewriting the template to exclude the "refrain from editing" portion and to align it a bit closer to the wording and tone of {{uw-coi-brief}} by tomorrow. When I finish that rewrite, I will crosspost to the AI noticeboard and ping y'all, unless you prefer me not to. Thank you guys for your help! - Otherwise (Antibacklog Aktion!) (talk?) 02:18, 5 August 2026 (UTC)

@Mathglot @Sammi Brie

I've changed the wording to remove the "refrain from editing" portion and to make it a bit closer in tone and structure to {{uw-coi-brief}}. I have also crossposted a link to this discussion to the AI noticeboard. Thanks! - Otherwise (Antibacklog Aktion!) (talk?) 15:09, 5 August 2026 (UTC)

Phrasing to avoid confusion due to temporary accounts

Mathglot, I agree about "logged in" on Template:Uw-editsummary and I've retained that, but I don't think we should change "registered" to "Wikipedia" because temporary accounts can't configure preferences so it's not "all Wikipedia accounts" that can use preferences. My intent was to mirror WP:ACCOUNT which uses "registered account" consistently. Also, using "a" rather than "your" avoids assuming that the reader already has a registered account since this template can land on temporary account talk pages too. Regards. Daniel Quinlan (talk) 20:55, 4 August 2026 (UTC)

Daniel Quinlan, yes, the fact that this template can be given to both registered and temporary account users is exactly the problem, because the wording doesn't work well for both without some unnecessary waffling to try to cover two dissimilar situations. Your most recent version is an improvement, but still doesn't quite resolve the problem.

I see two ways to do it: either say something like, "If you have a registered account and are logged in to it, you can..." or have specific wording for each case. I find the first alternative wordy, and still waffling. Why guess which one they are and have to cover both bases? It's trivial to detect a temporary account, so we should just do that and choose the right wording for a registered editor, and either no statement for the TAs, or an exhortation to register with a teaser about the feature. Perhaps something like these:

1. (registered): you can give yourself a reminder by setting Preferences → Editing → Editor → Tick Prompt me when entering a blank edit summary (or the default undo summary), and then click the "Save" button
2. (temporary): If you register for an account, you will be able to set a reminder to avoid leaving a blank edit summary (or nothing).

And it's always nice to have an opportunity to remind a temp user of the benefits of an account, and this is a very concrete reason to have one and is related to something they just did, so it's very relevant in this context. Thoughts? Mathglot (talk) 21:47, 4 August 2026 (UTC)

I prefer the current phrasing for how to set the reminder: you can give yourself a reminder by setting Preferences → Editing → Editor → Prompt me when entering a blank edit summary (or the default undo summary), and then click the "Save" button.

I would suggest varying the start of the sentence like If you are logged in to your account, ... for registered accounts vs. If you register an account and you are logged in to it, ... for temporary accounts.

Daniel Quinlan (talk) 22:06, 4 August 2026 (UTC)

Daniel, there's a sandbox version available for comparison at {{Uw-editsummary/sandbox}}. Note that a testcases page cannot test this feature, as it depends on the pagename of the page the template is placed on. You can test it at Special:ExpandTemplates, using the "Context title" to specify the pagename as the User talk page of either a temp account, or a registered account, and observe the differences. Alternatively, you can place it on real User talk pages, and use Show Preview, without saving. Lmk what you think. Mathglot (talk) 23:19, 4 August 2026 (UTC)

Change is in place. It's a subst: template and only used once every few hours so I just tested with Twinkle on two user pages that needed the reminder about edit summaries. Worked fine on both. Cheers. Daniel Quinlan (talk) 01:24, 5 August 2026 (UTC)

That was the right idea, but unfortunately it depends on {{is temporary account}} (c.b. {{if temporary account}}) which has more failing test cases than working ones. Example:

{{if temporary account|~2ABCD-barf-whoopsie!-giraffe|is a TA|is not a TA}} → is a TA

So {{is temporary account}} needs repair, because it tags almost everything as a TA. I have a Lua pattern that gets it right for every test case I can come up with, but I need to check with VPT tomorrow, because the definition given at WP:TA is wrong, and misses thousands of valid TAs. Will get back to this soon. Mathglot (talk) 04:21, 5 August 2026 (UTC)

@Mathglot, this test case isn't necessary in practice, as account names starting in ~2 can't be registered normally, meaning any such username will be a TA. Chaotic Enby (in solidarity · talk · contribs) 04:25, 5 August 2026 (UTC)

Oh, that's good to know. Is that documented someplace? As a separate issue: TA's def is wrong, and I will get to that later; afaict an exact def is given (for this year) by PCRE ^~2026\-\d{3,5}(?:-\d\d?)?$ (the first backslash is optional). My lua is weaker, but I believe this is not expressible in one Lua regex, but ^~2026\-\d{3,5}\-?\d?\d?$ is probably close enough for all practical purposes. Thanks, Mathglot (talk) 04:34, 5 August 2026 (UTC)

Yes, it is at Wikipedia:Username policy#Temporary accounts, which is why {{is temporary account}} is coded the way it is instead of going for a more elaborate (and expensive) regex that fits exactly the possible TA usernames. Chaotic Enby (in solidarity · talk · contribs) 04:36, 5 August 2026 (UTC)

Daniel, I restored your version based on Chaotic's explanation of ~2. Thanks, Mathglot (talk) 04:58, 5 August 2026 (UTC)