Steward requests/Miscellaneous/2026-03

Manual requests

Request to reverse the vanishing of User:Renamed user 8281b6079d3b1708114df3eee5f3a3d6

Status:    Done

• Renamed user 8281b6079d3b1708114df3eee5f3a3d6 [stalktoy] – [cross-wiki edits]→Special:Contributions/MisterTung

I have noticed that this user was indefinitely blocked on the grounds of being a sockpuppet of Fairyspit on enwiki. Therefore, I am making this request. Please see w:en:Wikipedia:Sockpuppet investigations/Fairyspit and Steward_requests/Global#Global_lock_for_Fairyspit_and_sockpuppets. Kindly unlock the account, or, after unlocking, relock it with the correct log entry if necessary. Thank you. --浅村しき (talk) 16:58, 4 March 2026 (UTC)

Done AramilFeraxa (talk) 08:56, 6 March 2026 (UTC)

dzwiki machine translations

Status:    Done

Please delete the following:

All of these were created and edited by the same LTA (there are also edits by other users, but those are trivial) using multiple socks:

NguoiDungKhongDinhDanh 02:34, 8 March 2026 (UTC)

Done --Min☠︎rax^«¦talk¦» 06:00, 8 March 2026 (UTC)

Add a CSS instruction on this page so that the scores do not exceed the width of the page

Status:    Done

Please add the following code to this page so that the scores do not exceed the width of the page on the breton Wikisource:

/* Na lakaat ar skridoù-sonerezh da vont en tu all da vent ar pajennoù [[phab:T67252]] */
.mw-ext-score img {
    max-width: 100%;
    height: auto;
}

--Gwendal (talk) 09:33, 8 March 2026 (UTC)

-𝐖𝐢𝐤𝐢𝐁𝐚𝐲𝐞𝐫 👤💬 (WikiBayerCatHelper) 10:00, 8 March 2026 (UTC)

eewiki edit requests

Status:    Done

Please apply the following changes to ee:Template:Unicode and ee:Template:IPA, correspondingly:

Extended content
Line 1: Line 1: <templatestyles src="Unicode/styles.css" /><span class="Unicode">{{{1}}}</span> <span class="Unicode">{{{1}}}</span> Line 1: Line 1: <templatestyles src="IPA/styles.css" /><span title="Pronunciation in IPA" class="IPA">{{{1}}}</span> <span title="Pronunciation in IPA" class="IPA">{{{1}}}</span>

These two are inline templates and so shouldn't have extraneous whitespace. NguoiDungKhongDinhDanh 00:12, 10 March 2026 (UTC)

Done — xaosflux ^Talk 01:28, 10 March 2026 (UTC)

@Xaosflux: ee:Template:IPA still has a leading space. NguoiDungKhongDinhDanh 01:39, 10 March 2026 (UTC)

done.🪶-TΛNBIRUZZΛMΛN (💬) 06:32, 10 March 2026 (UTC)

Big deletion of a sandbox

Status:    Done

Please delete ru:Участник:Qweasdqwe/Черновик. It's an old sandbox that should be speedily deleted since the user has been inactive for 2+ years (he passed away). But the page has more than 5000 versions, so an administrator was not able to delete it, and neither was I (being a closer, I have the 'delete' right, but not the 'bigdelete' one). It will get us rid of 3 LintErrors and of spending bots' time on keeping that page relevant and tidy. Shabe (talk) 21:59, 13 March 2026 (UTC)

Done A09|(pogovor) 22:02, 13 March 2026 (UTC)4

Create and configure a new wikiset

Status:    Not done

Per the guidance provided in phab:T417505#11618361 I would like to request a new wikiset to be created, similar to the existing wiki set titled "Opted-out of global bots". This new wikiset should be called "Opted-out of global IP block exemption" and should be an opt-out based set. As a start, fawiki should be listed as a member. Special:GlobalGroupPermissions/global-ipblock-exempt should then be modified to use this wikiset. Huji (talk) 02:53, 16 February 2026 (UTC)

For the record, there is a relevant discussion at SRGP/2025-11 § Global IP block exempt for GodNey (revoke). NguoiDungKhongDinhDanh 03:16, 16 February 2026 (UTC)

 Oppose GIPBE only bypass global IP blocks and rangeblocks, and it does not affect local IP blocks and local rangeblocks. 🪶-TΛNBIRUZZΛMΛN (💬) 03:30, 16 February 2026 (UTC)

"users who know English and can navigate Meta would have a higher likelihood of getting around proxy blocks by way of GIPBE" sounds out of touch: GIPBE is usually granted only to established users. Besides, local policy should not prevent blocking the proxy IPs used by sockpuppet suspects found during CU (as a way to force them apply for local IPBE). Is there local concensus for the change you propose? Srapoj (talk) 06:09, 16 February 2026 (UTC)

• I'm a bit hesitant to this, mainly because a) this right is already quite sparingly granted to users who are deemed trusted, and b) the purpose of this right is to be truly global to allow such users who have a need to edit from globally blocked IPs. If fawiki has a compelling reason to opt-out of this group, it should likely at the least go through a community discussion, similarly to how we require communities to start discussions for opting in/out of, for instance, global sysops. EPIC (talk) 06:29, 16 February 2026 (UTC)

  Oppose: This is a global right that by the global community. A single project cannot change this; it must be changed by a global RFC. 𝐖𝐢𝐤𝐢𝐁𝐚𝐲𝐞𝐫 👤💬 (WikiBayerCatHelper) 16:20, 16 February 2026 (UTC)

@Huji, what makes fawiki special enough that it needs to be excluded from GIPBE (I did review the phabricator thread and it does sound a bit odd that you're forcing these users to take risks)? Also I thought that you already block proxies like enwiki does. Leaderboard (talk) 16:43, 16 February 2026 (UTC)

A couple responses:

• Can local communities still block users from using proxies? Yes, but that means they need to import all proxy blocks locally. The purposes of global blocking proxies is, partly, to avoid every project having to initiate, extend, and manage thousands of proxies blocks.
• Why is fawiki special? It isn't special. However, because most editors in fawiki are from Iran (where many websites are blocked by the government) and most of them use proxies for other internet browsing, many of them find themselves blocked when accessing Wikipedia. It may be an inconvenience for them to turn off proxy when editing Wikipedia, but we want CU data to be reliable, therefore we expect them to accept the inconvenience, and we sparingly offer local IPBE. By issuing GIPBE, Stewards are ignoring the wishes of the local community.
  With that said, I would argue any project with active local IPBE process should be allowed to be exempted. Cross-wiki users can request local IPBE on those projects, allowing the local community to assign this right. A list exists in No_open_proxies#Local_exceptions but Stewards do not consult with or defer to those projects when issuing GIPBE.
• What if this proposal gets declined? Then the only remaining solution would be for a bot to constantly import proxy blocks from meta into fawiki.
• Do stewards issue GIPBE for its intended purpose (that is, for crosswiki editors who are stuck in IP blocks across many projects)? Not always. There are many cases of GIPBE being granted to a user who (aside form Commons), really only edits in one Wikipedia or a couple. Those users should be directed to the local IPBE processes of said wikis, but an attempt to convince Stewards to do so has also failed.

If Stewards neither agree to a wiki-based exemption, nor to a process in which they involve local communities and limit GIPBE to truly cross-wiki editors, then I would like that to be made known here so I can pursue alternatives through phab:T417505 and/or mass importing global blocks into local wikis. Huji (talk) 18:06, 16 February 2026 (UTC)

Neutral. On one side, I don't want to limit the right of a project to do what it wants, and if it involves GIPBE, so be it. On the other side, I still think this is a bit heavy-handed (after all, you need to think from the user's point of view - why should they have to unnecessarily apply for rights on multiple projects? A criticism also applicable to en.wikipedia), but again not my call to take. Leaderboard (talk) 19:16, 16 February 2026 (UTC)

• I don't really understand why a project would trust stewards to effectively globally block open proxies, but not effectively grant exemptions to those global blocks on an as-needed basis. All options presented are concerning absent information on the problem - have there been a number of cases where non-trusted users were granted GIPBE? If so, maybe stewards should be more restrictive in who is granted those permissions, particularly if those users are active on fawiki. Do fawiki admins lack capacity to deal with vandalism and abuse? Out of approx. 1250 users with GIPBE, presumably only a handful are active on fawiki - are local admins not able to block problematic users with GIPBE? If this is a case of one user being granted GIPBE who misused the permission, that doesn't seem to build a case for fundamental change to me. – Ajraddatz (talk) 22:13, 16 February 2026 (UTC)

  Great questions, Ajraddatz. The reason for the first part is that blocking proxies is an agreed upon process on all wikis (and Stewards doing it centrally makes sense), but deciding to whom to give the exception is a subjective decision. One is agreeable by all, the other is debatable.

  The answer to your second point about whether fawiki admins can block vandals is: the concern isn't with overtly vandal users. The concern is with the user type who is a sockmaster, causing conflicts without crossing the line. If we establish they are a sock via CU, we can block them. If the CU data is tainted because they were allowed to use proxies, we cannot. That is why we don't want GIPBE to be granted to users who aren't truly cross-wiki (i.e., doing hundreds or thousands of edits across many different projects). But, as I stated before, an attempt to create a guideline for Stewards to do exactly what you said ("be more restrictive in who is granted those permissions") was tried and failed.

  Not to make this political, but one could argue the Stewards have no skin in the game if a sock master causes conflict in a project on which they are not active. But Stewards could feel like their power is being taken away by way of a restrictive guideline or a restrictive wikiset config. This imbalance may explain why every effort to try to limit the GIPBE process fails. Huji (talk) 22:49, 16 February 2026 (UTC)

• Appreciate the follow up, and I can understand your frustration given you have raised the issue before. I am still a bit worried about the implications of stewards being able to issue a global block but not able to exempt people from it - particularly if multiple wikis are added to the opt-out set. It would create an onerous series of bureaucratic hoops for a user impacted by a global block to go through, and since the vast majority of users with GIPBE are not disruptive, I am not sure of the benefit vs the cost. I'm also not sure that your hands are as tied as you think they are in cases where someone is being disruptive but not a CU confirmed sock - you can still block or warn them for disruption, and ultimately there are easier ways for someone intent on causing disruption to mask their true IP address than editing through a proxy and requesting GIPBE.

All that said, I do think that stewards could be more restrictive in handing out GIPBE in cases where the requester is only active on one wiki. I'm not sure I agree with your thresholds below - 500 edits is more than most accounts make overall and Wikidata and Commons are two projects where people are typically somewhat active in addition to their home project. But the point is taken that a lot of users end up with GIPBE without having a significant history of cross-wiki editing. I will try to think more about that myself in my own GIPBE granting, and encourage other stewards to do the same. – Ajraddatz (talk) 18:49, 17 February 2026 (UTC)

@Ajraddatz I appreciate your thoughtful response, your willingness to understand a different point of view, and your commitment to change your approach and encouragement to others to do the same.

No doubt Wikidata and Commons are two projects where people are typically somewhat active on besides their home project. But that is the reason I excluded the data in my analysis below. Otherwise, a user who is highly active in 3 Wikipedias and a user who is highly active in 1 Wikipedia + Commons + Wikidata would appear similar. I am open to a different analytical approach too. Perhaps I should do the analysis by separating out and grouping together all wikis with local IPBE from all others; after all, one could argue for wikis with local IPBE the best approach is to drive the users towards that process, and reserve GIPBE for other wikis. Huji (talk) 00:39, 18 February 2026 (UTC)

As a follow up, I did some analysis. Turns out more than half of users with GIPBE are only really active in a single wiki (defined as having more than 500 edits on the wiki, excluding commons and wikidata), and another 20% of them are really only active on 2 wikis (with enwiki being a common second wiki for them). Most of these are zhwiki users. In other words, in more than half the cases, Stewards are granting GIPBE to users who are primarily and almost exclusively active on a single wiki that typically has its own internal IPBE. Why not let those communities grant local IPBE instead? Please be sure to check the graphs at the bottom of my analysis. Huji (talk) 00:57, 17 February 2026 (UTC)

@Huji: Your Jupyter notebooks are not visible to others (or at least not to me). NguoiDungKhongDinhDanh 00:59, 17 February 2026 (UTC)

@NguoiDungKhongDinhDanh: public link. – DreamRimmer ■ 01:07, 17 February 2026 (UTC)

• Oppose, reason being that this is a tool that stewards use to balance their own blocks - so they shouldn't be prevented from allowing a bypass of their own blocks. If a local project wants to block something, they can do it themselves. — xaosflux ^Talk 01:33, 18 February 2026 (UTC)
•  Strong oppose: On what grounds?? First of all, the fact that we GIPBEers are granted this privilege already shows that we are cross-wiki active users with no obvious misconduct. When we edit on fawiki, we are commonly there to counter vandalism — so why add another barrier that interferes with cross-wiki active users like us? Secondly, since we have GIPBE, we are legally allowed to use proxies to edit all sites, and in fact we need to use them. So why impose an additional obstacle on us? A simple local block would suffice. Why should GIPBE be rendered completely ineffective on a particular site? If this wikiset is established and a bunch of sites are added, then what use is GIPBE to us anymore? I feel that instead of making this proposal, it would be better to propose abolishing the GIPBE user group altogether. 浅村しき (talk) 15:36, 4 March 2026 (UTC)
•  Support None of the people here should have the right to second-guess the way fawiki wants to structure its user rights, even if you think that they are silly and misguided. * Pppery * _{it has begun} 04:59, 6 March 2026 (UTC)

  What? No part of this has anything to do with local user rights. That local project is welcome to manage all of their users, their local rights, and their blocks. No part of this has anything to do with overriding local governance. — xaosflux ^Talk 13:46, 6 March 2026 (UTC)

  On the contrary, the decision "only local admins can decide who is allowed to edit through proxies" is precisely a local governance decision that, as I see it, you are unfairly vetoing. * Pppery * _{it has begun} 17:04, 6 March 2026 (UTC)

  Local projects may block all the proxies, or any other IP or IP range they would like to, and once they do they have direct control of who may bypass their blocks. Additionally, local projects have the ability to override or bypass external blocks that affect their project, either entirely or for their users or usergroups. — xaosflux ^Talk 19:19, 6 March 2026 (UTC)

•  Oppose per xaosflux's stated reasoning. Codename Noreste (talk • contribs) 19:14, 6 March 2026 (UTC)
•  Oppose Xaosflux is right here. Local wikis wanting more control over proxy blocks are welcome to simply adopt all proxy blocks locally. Nothing is stopping them from locally blocking every identified proxy, so if it's such a big deal for fawiki, surely some local admins would be more than happy to do so. EggRoll97 (talk) 05:59, 7 March 2026 (UTC)

I close this request per discussion.--𝐖𝐢𝐤𝐢𝐁𝐚𝐲𝐞𝐫 👤💬 (WikiBayerCatHelper) 18:55, 8 March 2026 (UTC)

Going back to in process as status. Such requests should only be closed by stewards. As Amanda said above, we (stewards) will discuss this. I understand, that the discussion here seems to have some consensus, however, determaining consensus is up to the stewards. Furthermore, you've commented on this already and shouldn't close requests you are involved with. -Barras talk 19:33, 8 March 2026 (UTC)

Not done Closing as not done after over 1 month of discussion. There is clear consensus against the creation of this wikiset. shb (t • c) 04:01, 18 March 2026 (UTC)

Creations by Yasmin taha at elwikibooks

Status:    Done

Please delete these as likely copyvios:

• b:el:Βιβλίο μαγειρικής:Κοτόσουπα (significantly similar to this post, but there are a few differences; plus, the article was from 2021 and that post was from 2023)
• b:el:Βιβλίο μαγειρικής:Σαγανάκι (similar to this post, but only partially)
• b:el:Βιβλίο μαγειρικής:Μπριάμ (significantly similar to this post, but the post has no publish date)

The same user created 7 pages. I already tagged 4 of them for deletion as copyvios. While there are not enough evidences to definitively say that these were also copyvios, they most likely are. NguoiDungKhongDinhDanh 03:49, 14 March 2026 (UTC)

Done Large portions of the content on these pages were indeed copyright violations, so I have deleted them. – DreamRimmer ■ 12:36, 18 March 2026 (UTC)

Unvanishing of User:Orbovia

Status:    Done

Orbovia, now User:Renamed user c0fa4c6665710e25e3366c5f2b37e05e, has abused the en:WP:VANISH process to evade scrutiny and continue with disruption that volunteer editors have to fix. See en:Wikipedia:Administrators' noticeboard/Incidents#Edit warring. They have been LOUTSOCK-ing after their "vanish" (which requires the user leave the project) and disrupting Talk pages, showing it was only for avoiding scrutiny and not a true vanish. An administrator and several users at enwiki ANI have agreed and the relevant IP has been blocked for socking, but the issue of them continuing to edit (and they may further edit by avoiding the new IP block) despite vanishing to avoid scrutiny that was happening to them at the ANI remains. This shows that they have abused the process, and it should be reversed per "If the user returns, the vanishing will likely be fully reversed..." in the policy text. As the evidence involves non-public info that requires tools (including TAIV), I won't publish it publicly but if any Stewards want to see more than what's said at the ANI (which should be enough, I think, but can clarify if needed), I can email them on request. This is a Meta/Steward issue as those perms are required to reverse a vanish. Thank you. MolecularPilot (talk) 08:55, 22 March 2026 (UTC)

Seconded, this editor explicitly confirmed to me on their Talk page that they wished to vanish during a dispute with another editor.

Within minutes, a temporary account appeared to reinstate their disputed edit. When this was reverted, a second TA appeared to do the same thing.

This is clear and blatant misuse of the courtesy vanishing process, as the editor did not wish to leave Wikipedia permanently and intended to continue their dispute without any link to their current account.

They were made fully aware that vanishing involves permanently leaving the project, as I sent them the instruction page that has it prominently and clearly written at the top.

Diffs are in the ANI thread but I'll be happy to provide here if required. Blue-Sonnet (talk) 13:10, 22 March 2026 (UTC)

@MolecularPilot: Done A09|(pogovor) 16:15, 22 March 2026 (UTC)

Thank you! Have a nice day. :) MolecularPilot (talk) 21:46, 22 March 2026 (UTC)

Delete bot sandbox page

Status:    Done

Hello! Can someone please delete this page? It's my bot's sandbox page on the Albanian Wikipedia. I recently updated its code and the page is not needed anymore but given its large history, admin rights are not enough to delete it apparently. - Klein Muçi (talk) 19:31, 24 March 2026 (UTC)

Done Barras talk 19:34, 24 March 2026 (UTC)

~2026-18668-77

Status:    Not done

Please delete all userpages of ~2026-18668-77 across GS wikis as vandalism. Pages like these inflate the statistics. NguoiDungKhongDinhDanh 05:08, 28 March 2026 (UTC)

Not done I am declining this request as there is no need. The effect on any statistics is miniscule. Count Count (talk) 14:47, 28 March 2026 (UTC)

Mass machine translations/AI-generated pages

Status:    Done

Jon Gua (talk · contribs · CA) has been mass-creating AI-generated articles or at the bare minimum machine translations as they are adding many pages to many many wikis where they do not know the language. They have been blocked for this on simplewiki as a few days ago (by myself) but are still continuing identical behaviour on many other wikis. Can you review to see if any action is necessary? Thank you! --Ferien (talk) 18:44, 17 March 2026 (UTC)

How can you be certain that I don’t know the languages I'm editing in? That's a serious assumption, and I haven’t seen any evidence provided to support it. If there are specific concerns about the quality or accuracy of my contributions, I'm open to reviewing them and improving where needed, and I'm happy to discuss any particular edits that may be problematic. But broad claims about my language ability or intentions aren't helpful without concrete examples. Jon Gua (talk) 21:05, 17 March 2026 (UTC)

Jon Gua, ok, here's evidence. You have been creating pages in English (through simplewiki), Interlingua, Ancient Greek, Extremaduran, Ladin, Esperanto, Papiamento, Afrikkans, Ido, Spanish, Lingua Franca Nova, Latin and Galician, at least, as well as a variety of other languages on Incubator. Of course one could assume that you just speak all these languages really well, but there is evidence otherwise. It is more difficult to identify AI-generated content in languages I don't speak, but you can definitely still find the AI-generated vibe in your articles, and it backs up too: On eowiki you claim to be eo-2, yet you added to the article eo:Artaj bildigoj de Viriato:

Arta bildigo de Viriato reflektas evoluantajn naciajn rakontojn en la Iberia Duoninsulo. En Portugalio, li ofte estas akceptata kiel pra-nacia heroo antaŭ la romiigo de la regiono. En Hispanio, li same estas honorata en luzitaninfluitaj areoj kiel en parto de Ekstremaduro. En ĉiuj formoj, lia bildo enkorpigas temojn de libereco, rezisto, perfido kaj la komplikecoj de kolonia historio.

Relying on machine translation back to English, you get

Artistic depictions of Viriato reflect evolving national narratives in the Iberian Peninsula. In Portugal, he is often accepted as a proto-national hero before the Romanization of the region. In Spain, he is also venerated in Lusitanian-influenced areas such as in parts of Extremadura. In all forms, his image embodies themes of freedom, resistance, betrayal, and the complexities of colonial history.

This sounds an awful lot like the overgeneralisation and undue emphasis on legacy that featured in many of your simplewiki articles that is a classic sign of AI-generated content. And it's such a good sign of AI-generated content because it says pretty much nothing of significance yet is somehow incredibly advanced/complex at the same time, certainly beyond level 2 that you claim to be. This also applies for pap:Deskripshonnan artístiko di Viriato, which, upon review, seems very similar to simple:Artistic depictions of Viriathus, which has many clear AI-generated signs I can easily identify in English. --Ferien (talk) 01:36, 18 March 2026 (UTC)

Thank you for providing specific examples. If I understand correctly, part of your concern is based on the language levels listed on my user page. In that case, I would like to clarify that those levels are self-assessed. I haven't taken formal exams for most of these languages (for some of them they don't even exist), so I chose an approximate level (often "2" as it is the mid-point) as a rough estimate rather than a precise measure.

Regarding the number of languages: many of the ones you mention are Romance languages. Given that my native language is also Romance, it's not unusual to have reading or writing ability across several of them, especially for general topics. Similarly, auxiliary languages are intentionally designed to be easy to learn, and it's quite common for contributors in those communities to work across multiple such languages.

For Incubator work (and some of my work across several Wikipedias, not all of them), I'd also like to clarify that I'm often not the one translating. I frequently help with templates, categories, and structure, and when needed I ask speakers of the language to assist with or review content.

On your point about "overgeneralisation": I understand that this can sometimes be associated with AI-generated text. However, the presence of that style alone is not evidence of AI use. If there are specific inaccuracies, mistranslations, or policy issues in particular edits, I'm more than willing to review and correct them.

Finally, regarding similarities between articles across languages: adapting or translating content between wikis is a long-standing and normal practice, so I don't see any issue with that.

If there are concrete issues with specific edits, I'd appreciate it if you could point them out so I can address them directly. --Jon Gua (talk) 06:41, 18 March 2026 (UTC)

@Jon Gua: Can you clarify whether you've used AI to write that response above? I ask because of this (a 100 per cent ZeroGPT match). //shb (t • c) 06:51, 18 March 2026 (UTC)

Sorry, but can we focus on topic of the conversation? AI-detection tools produce false positives, so that is not really a proof of anything. I repeat myself: I don't mind addressing any issues regarding my contributions like the translation quality or problems with references as I believe focusing on concrete examples will be more productive than using automated detection tools. Jon Gua (talk) 08:51, 18 March 2026 (UTC)

You have utterly failed to address my point. Did you use AI in your reply to Ferien, yes or no. That's all I'm asking for. I don't want deflections, I want a concrete answer. //shb (t • c) 08:56, 18 March 2026 (UTC)

No. But again, I believe it is more productive to focus on my contributions not on the conversation itself. Jon Gua (talk) 09:04, 18 March 2026 (UTC)

Thank you SHB for noticing that these responses read as though they are AI-generated as well. @Jon Gua: For Incubator work, you actually are often the one translating - I am not quite sure why you would say this when your last article in Ancient Greek was created just 8 days ago, and based on this (only including pages created in Mainspace), you've been creating articles in grc, sas, ssy, aig and wlx in the first 50 only.

Indeed you make a fair point that that article has mainly one sign of AI that I noticed from you in the past, because I haven't been scrutinising every page - again, it is quite difficult to - but as far as the signs go from you, I don't think I've ever seen one editor have so many pages that consistently have AI-generated features as well as such a variation of those features. There is actually no global rule on AI - maybe there should be soon? - but from my perspective, if you can't tell an article is AI-generated, it is probably ok, this is not the case here though. The overgeneralisation that I've mentioned is a completely concrete issue, it means these sections prove difficult to cite because the claims are sweeping. Why are you thanking me for providing specific examples at first but then by the end of your response asking me to point out any specific issues if any exist, when I have already done so once (and therefore, there exists an issue I have pointed out)? Just one example of how your use of AI is also making discussions difficult and so definitely worth pointing out. I will see if I can investigate into this further at some point later this week, likely tomorrow. --Ferien (talk) 09:52, 18 March 2026 (UTC)

Thank your for your detailed explanation. I want to clarify that the GRC articles are revised and reviewed by another editor when they have time, as you can see by the templates and categories created to address that concern (there are categories for articles that need to be revised, articles under review and revised entries.

As I mentioned you on our conversation at the Simple English Wikipedia, I now understand that the issue is not the use of AI, but complex language and overgeneralisation. I may have misunderstood the policy before. What I have been doing is summarising and translating existing articles, then rewriting them in what I thought it was Simple English. Clearly, some of those are not simple enough.

Again, I am willing to revise and simplify any articles that do not comply with the Simple English guidelines to addres both language complexity and overgeneralisation. Jon Gua (talk) 10:12, 18 March 2026 (UTC)

"I now understand that the issue is not the use of AI, but complex language and overgeneralisation." – complex language on simplewiki is a problem, yes, but AI-generated content is also just as much of a problem. AI can hallucinate and AI-generated content can just be pure fluff and nonsense, sometimes also with fake references. //shb (t • c) 10:33, 18 March 2026 (UTC)

I totally agree about AI hallucinations, but, as I said, my articles are most of them translations from already-existing articles, so I avoid that issue completely. Jon Gua (talk) 10:45, 18 March 2026 (UTC)

┌─────────────────────────────────┘
Apologies for the delay on more complete evidence. I'd like to show the following examples:

• w:tum:Wendy Ashby - yet another language written in to a high level that I had not initially included, and not a Romance or auxillary language. The article as a whole features quite complex language and a non-standard list, though the section Kukolerana kwa Luso: Qo Akti? stands out more. As far as I can tell from the version I have machine translated into English, it features a negative parallelism, the rule of three and an undue emphasis on legacy:

  In addition to her linguistics work, Wendy Ashby has been involved in a multidisciplinary art project with German artist Mathilde ter Heijne, contributing to the film Qo Akti?. This experimental film, part of a broader exploration of feminist culture and language, involved Glosa translating her text, including language in the framework of art theory. Ashby’s involvement not only brought Glosa to contemporary art, but also demonstrated the versatility of the language in terms of narrative and metaphor. The project was presented as part of ter Heijne’s wider project on culture, female sexuality, and solidarity.

• w:zh:温迪·阿什比 - created last week, another translation of Wendy Ashby. Already identified as a poor translation on zhwiki. The format is clearly similar to the tumwiki article shown above (you don't need to be a speaker of either language to identify that) which was already at a high level of complexity. +1 to the number of languages, Mandarin which is by no means comparable to a Romance or auxillary language.
• w:io:La Celestina - this one stood out as one that had a lot of different vocabulary.
• w:ext:Malpartía de Caçris - the following ones are much more revealing! In w:ext:Special:Diff/132514, a Spanish section alongside the Extremaduran is added (and rapidly removed after creation), that roughly translates to:

  The cuisine of Malpartida doesn't differ too much from typical regional cuisine, but some of the most traditional dishes include Chanfaina (a type of offal stew), Berzas (a cabbage stew), fried tench, and lamb stew. As for pastries, Roscas de Aire (a type of donut), Roscas Fritas (fried donuts), and Roscas del Calvario (a type of donut) are worth mentioning.

"worth mentioning"? This could either be an AI prompt from Jon Gua himself or AI prompting Jon Gua or what to add. Either way, there is no way such Spanish would appear in any article normally.

• w:pap:Kanon oksidental - up until w:pap:Special:Diff/160121, the article features a duplicative sentence that uses the same reference 3. Running along the translation reasoning Jon provided, that English writing Another critique highlights a narrow interpretation of the West, dominated by British and American culture, at least under contemporary circumstances, prompting demands for a more diversified canon amongst the hemisphere. actually makes no appearance in the equivalent English Wikipedia article and, from my perspective, appears AI-generated.
• w:ia:Germania - and the most conclusive of all, lines from an LLM appear directly on the text in w:ia:Special:Diff/695248.

*Wait, "Jews":* `judeos`.

*Wait, "Jehovah's Witnesses":* `Testimones de Jehovah`.

*Wait, "Ukrainians":* `ukrainianos`.

*Wait, "Belarusians":* `bielorussos`.

*Wait, "Polish people":* `poloneses`.

*Final check on the text:*

All articles feature the rough signs of non-existent categories and see also sections with redlinks. So not only do we have the sheer number of languages involved and telltale signs but we also have some accidental copy and pastes from AI in there now too. --Ferien (talk) 22:50, 21 March 2026 (UTC)

I will try to answer everything one by one:

• Tumbuka: I actually know the administrator and we worked together in developing the Wikipedia, so nothing done really in isolation.
• Ido: That one is correct, so I don't know what the issue is. Different vocabulary? I do speak Ido, so I don't need to translate exactly word by word every single article I create, they are my own wording. Definitely I didn't use any translation machine nor AI for this one (I don't think it even exists).
• Extremaduran: There is no translation machine nor AI that can translate to Extremaduran. I did it myself. When I published that edition, I forgot to delete the Spanish text I was translating it from, an edition mistake I amended by deleting it. There is nothing wrong with that. BTW, the text comes exactly from the Spanish Wikipedia. It is still there.
• Interlingua: I write the articles on Word first, and sometimes just "correct" automatically some words (I notice that when revising some articles (like when changing the centuries or the BC acronym). In some other cases, sometimes there might be two correct options, sometimes could be my own mistake (quite common when writing in constructed languages). That is why there is another user revising the articles.
• Regarding categories, I usually translate the ones from the original article because I don't really know what to do. I don't know whether it is better to leave them in red until there are a few articles in there and then create the category, or maybe put them in another category that is not the same as in the original article. Because I don't know, I tend to leave them like that. Maybe it is wrong, I don't know.
• I hope I have addressed the issues that you have mentioned, yet I want to point out the fact that, according to your user page, you don't speak any Romance language nor any constructed language, yet, you are accusing me of using the wrong language. If you don't speak Ido, Extremaduran or Interlingua, how come you point out "mistakes"? How can you know there is any mistake if you don't speak the language? It looks that you are just looking at "what seems wrong" trying to find something to accuse me, instead of really looking into it. Just trial and error to see if you might maybe tap into something, which I think goes against AGF policy. The example of the Extremaduran Wikipedia is very clear, as it even appears today as it was in the original. As I said before, the fact that "looks like AI-created" doesn't mean it has been done by using AI, as proven by the Extremaduran example. The fact that you dedicate time to try to find "anything wrong" in my contributions when you don't even speak the languages you are monitoring to try to find mistakes is a proof of not assuming good faith, especially when you didn't even compare it with the original article!
• I repeat, if I made mistakes at the Simple English Wikipedia regarding overgeneralisation and complexity, I am willing to revise and amend them. I do also understand that sometimes, some of my contributions can look like AI-generated but again, that could be a result of me translating and summarising already-existing content.

Jon Gua (talk) 08:16, 22 March 2026 (UTC)

For w:ext:Malpartía de Caçris, I will strike that from the evidence as it is not clear evidence of AI use.

On w:ia:Germania however, you are the only human editor to have edited the article since October 2022 (as you have admitted to being the same person as Caro de Segeda, yet did not admit that on the many wikis you have edited) and expanded it from 3,571 bytes to a staggering 107,814 bytes entirely independently, and you have not engaged in any on-wiki discussion outside your own talk page or welcoming since December 2025- so as for another user revising the articles, I find no evidence of that.

It is important to say that assuming good faith is not about assuming good faith when evidence to the contrary pops up. The evidence about AI-created articles is frankly just a bonus. You've created articles in 116 languages (including the Incubator languages). To write an article in any language, you need quite a bit more beyond basic comprehension, especially to go into the level of detail you have been going into. And we could go around in circles again how some of the languages are similar, some of them are variants of each other, not all the articles go into exceptional detail, but

• the number involved would still be truly extraordinary for a human to even have understanding at this level yet the actual quality of the articles is anything but,
• a wild number of AI-generated signs in simplewiki in English (you claim to be en-4),
• though this was more difficult to prove those signs have been reflected in many of your other language creations,
• despite this, text clearly generated by AI clear evidence was found as you had copy and pasted too much from the AI response to also include its writing towards you,
• AI-generated responses to these very concerns,
• the moment SHB pointed out your second response had a score of 100% on GPTZero, all your responses after this had scores of 0,
• extracting issues from this interaction that were simply never mentioned or remotely applied I now understand that the issue is not the use of AI, but complex language and overgeneralisation., Regarding the number of languages: many of the ones you mention are Romance languages. (given the number of languages involved that is an exceptionally poor excuse), If there are concrete issues with specific edits, I'd appreciate it if you could point them out so I can address them directly. (following a concrete example being bought up), etc,
  • the Romance and auxiliary language excuse falling apart is worth noting as a point on its own as you had created w:zh:温迪·阿什比 less than 4 days before that comment, it was tagged as a poor translation the following day, and this continues with the creation of w:is:Interglossa just 3 days before the comment and the creation of w:sv:Glosa (planspråk) just 2 days before the comment,

and need I go on? I am not going to find any more evidence just to keep going until you finally admit to using AI or machine translation as has been clear from the sheer number of languages involved. And this is not assuming bad faith. I first messaged you on simplewiki about AI-generated articles in December 2025 so this is not a new concern, in that case it was not a matter of looking at what "seemed wrong" because your articles actually were nowhere near in scope of the project as they were so complex. Going through the pages you have created has actually been harder in this respect because as you mentioned, I don't speak the many languages at hand so seeing things that are wrong has been more difficult, so yes, I do have to find what seems wrong, though I find this odd to point out when you do not know 116 languages either. If a steward or global sysop does not believe there is sufficient evidence, however, they are welcome to ping me and I will find more. --Ferien (talk) 21:12, 22 March 2026 (UTC)

With the amount of evidence in this thread, it's clear you [Jon Gua] have used AI to write a substantial amount of articles, and your responses are not convincing enough. Your excuses seem to be unusually niche and specific that it is more likely that you used AI than not. As such, I'll be deleting all your articles on small wikis without active sysops in 24 hours unless anyone reasonably objects. //shb (t • c) 10:07, 24 March 2026 (UTC)

Regarding the AI accusations:

• Most of my contributions are templates and categories, which I explained they don't need assistance at all (for instance: from the article about "geography" on X language, I create the category "geography", no need for any translation machine or AI).
• Specifc examples used to prove that they were AI-created have been shown to be direct translations or editing mistakes (such as the Extremaduran example being a direct translation from the Spanish one and the Swedish article being further expanded by other editors).
• At the Simple English Wikipedia, the concern was complexity not even AI-use and I have repeatedly stated that I am willing to revise and simplify every single article I created.
• Claims on AI-use based on the number of Wikipedia projects I have helped in are based on assumptions without verifiable evidence.

I request tht any action regarding detelion focuses on verifiable issues with specific articles, not assumptions. Again, I confirm I am willing to revise and redo any article that doesn't meet the standards. Jon Gua (talk) 10:17, 24 March 2026 (UTC)

There is so much evidence laid out in this thread by Ferien that I am utterly unconvinced by your explanations that entirely miss the point or fail to address the points in question. //shb (t • c) 10:23, 24 March 2026 (UTC)

I understand your position, but I have shown that the supposed "evidence" comes from either misundertanding, partial context or misinterpretation. I have addressed every example that has been given and demonstrated it didn't involve AI so I am not sure about the evidence you are talking about. Could you please be more concrete? Jon Gua (talk) 10:28, 24 March 2026 (UTC)

Again, I will try to answer everything because it is a lot to answer (thanks for taking your time to gather all that information):

• I know I have contributed in a lot of Wikipedias (something unusual) but my contributions vary in all of them, so it will be very difficult to address all of them one by one. In most Wikipedias (including the Incubator) I just create simple templates and add categories, two actions that require very limited language knowledge. In other cases, I rely on existing articles, translations or assistance from speakers.
  • Regarding your example of Swedish: the article hasn't been deleted, indeed some people have expanded it, which suggests that the issue is not that clear, can you please clarify the specific concern with that example?
  • Another example you provided, the Extremaduran one, makes me concerned because you just derived your conclusions from incomplete context. This suggest that your conclusions might be taken from partial information. I also want to note that, once clarified, this example was removed from the evidence. Given that this conversation was originally focused on my contributions to the Simple English Wikipedia, I would appreciate if we focus on that so any concern can be addressed directly.
• I am also concerned with your statement I am not going to find any more evidence just to keep going until you finally admit to using AI or machine translation as has been clear from the sheer number of languages involved., because it suggests that you have already reached a conclusion. I believe it would be more constructive to focus on specific issues based on complete context and full information, not just assumptions.
  • My understanding regarding the the AI-issue is that there was a shift in the conversation regarding my contributions to the Simple English Wikipedia: we are talking here about complexity not the use of AI. Given the template and the lack of communication, there was a misundestanding from my behalf. I do recognise that my contributions there do not meet the standards and, again, I am willing to redo every single article that needs redoing.
• Again, I do understand that sometimes my articles may seem as created by using AI due to the fact that I summarise them and that I have created articles in many Wikipedias. I want to repeat that most contributions are not really language-related, are just templates and categories. For longer articles, I rely on assistance from speakers either inside or outside Wikipedia.

Jon Gua (talk) 10:10, 24 March 2026 (UTC)

Yet another page showing communication intended for the user from your other account. Visible in logs as it will likely be deleted shortly. --Ferien (talk) 23:04, 24 March 2026 (UTC)

Hi, when I click that link it appeared the message: One revision of this difference (10652614) was not found. This is usually caused by following an outdated diff link to a page that has been deleted. Details can be found in the deletion log. I don;t realy understand it nor what you intented to show me here. Jon Gua (talk) 05:41, 25 March 2026 (UTC)

As I said, it was going to be (and has now been) deleted quickly, because AI-generated articles are quickly deleted on simplewiki once they have received endorsement. That's why I linked to the log of when you created the page, that link was technically broken but I've now fixed it. It says: Here’s a simplified, Wikipedia-ready version of your article, keeping all sections and references intact, written in plain English. You can copy-paste it directly, and it uses the `<ref></ref>` system as requested: --Ferien (talk) 10:41, 25 March 2026 (UTC)

Jon Gua, please answer this question directly: have you or have you not used LLMs to create content on Wikipedia? Vermont (🐿️—🏳️‍🌈) 23:38, 24 March 2026 (UTC)

@Jon Gua: I missed that SHB has asked you a similarly direct question above. Consider this your second opportunity to answer correctly. Keep in mind that we have seen this edit, this edit, that you have created pages in well over a dozen languages, and that your edits have timestamps. Consider further that the sole question I have asked is whether you have used LLMs to create content, and not about the quality of this content. Regards, Vermont (🐿️—🏳️‍🌈) 23:59, 24 March 2026 (UTC)

As far as I know, there is not AI that can translate to Interlingua, so it is not possible to do that, therefore, it is not applicable.

Regarding the Interlingua example, the text inside [[ ]] is just untranslated links from source articles. I work with original text from different Wikipedias (mostly English, French, Italian, Spanish and Portuguese) so I may initially keep those links to then verify against the Interlingua Wikipedia, i.e. I can copy [[this unstranslated part|---]] and paste it on the original Wikipedia, check then whether that article has an interwiki to the Interlingua Wikipedia; if it exists, I translate the exact title as sometimes it is not an exact translation: if it doesn't exist, I decide whether to create an internal link or not. This is part of the process of creating, publishing, revising and correcting an article so it is just a formatting issue. Jon Gua (talk) 06:05, 25 March 2026 (UTC)

Sorry, one more point: I am a bit confused about the scope of this conversation as it started with an issue regarding my contributions at the Simple ENglish Wikipedia, which was explained to me that the matter was language complexity not AI use. I addressed those concerns, explained my way of working there and expressed my willingness to revise any articles that don't meet the guidelines. Since then, this discussion seems to have shifted to AI use across different Wikipedias and I have answered to the examples that were raised. Could somebody clarify the current verifiable concern we are discussing about? From my point of view, the issue at the Simple English Wiipedia is solved, yet I am still blocked. --Jon Gua (talk) 06:28, 25 March 2026 (UTC)

In one sentence it is that you have used AI as an alternative to machine translations to create pages crosswiki where you don't know the language (bolding for the avoidance of doubt as you are still stuck on the simplewiki block), and this has likely occurred across hundreds if not thousands of pages. But the more I discover and the more you dodge the point the more I start to think that this case is really testing the limits of SRM and probably needs a discussion much wider in scope, because you've been writing poor articles in languages you don't know since before 2009 and 17 years later you still don't get the point.

For the sake of checking whether your excuses are adding up yet/fun at this point, I went to ChatGPT to ask it to translate a random English Wikipedia article into Interlingua and it complied immediately. The untranslated links have, again, not been discussed here. --Ferien (talk) 19:27, 25 March 2026 (UTC)

At this point, the discussion has moved far beyond the original issue, which is my block on the Simple English Wikipedia. I am focusing on that because it is the matter that needs to be resolved.

As I explained earlier, my understanding is that the concern there was language complexity, not AI use (that is what you said). Given that this is the case, I have repeatedly express my willingness to review, simplify, and improve any articles that do not meet simple-enough standard. That offer still stands.

However, the discussion has changed towards broader claims about my contributions across multiple projects, including speculative conclusions and examples not directly related to the original issue. I have answered to those examples, yet you continue to bring new ones, making it almost impossible to solve the matter of this discussion constructively.

Just to be clear I don't create articles using AI. My contributions are basically translated and adapted already-existing articles from other Wikipedias. In most cases, they are just things that don't require language knowledge as templates, categories, and formatting. When needed, I ask for help from people from that specific Wikipedia or from people from other Wikipedias that speak the language.

In order to move this conversation forward in a constructive manner, I will address points related to the original topic of this conversation.

Given this, I would like to return to the original issue. I have acknowledge the fact that my articles might be too complex and I express my willingness to fix all of them. On that basis, I request that my block on the Simple English Wikipedia to be reconsidered so that I can address any remaining issues directly. Jon Gua (talk) 21:05, 25 March 2026 (UTC)

Put it bluntly, this discussion was never about your simplewiki block. I don't get what you're trying to achieve with your repeated deflections and your refusal to accept the problem at hand. //shb (t • c) 23:31, 25 March 2026 (UTC)

This conversation originated from my block at the Simple English Wikipedia, which is what I have been focusing on. That is the only matter that I am trying to solve here.,

As I explained earlier, I wasn't given a proper explanation of the block and once a clarification was provided, the concern identified was language complexity. I repeteadly offered to review and simplify any article that don't meet the requirements, an offer that remains opened and to which I found no answer to it. My intention is to solve this issue constructively.

I would also like to note under normal Wikipedia practice, blocks are preventive and when an editor shows willingness to address the issue, there should be an opportunity to solve it.

I have also answered to every example raised from other projects and even explained my method of working including how the fact that I tend to summarise and reword articles from other Wikipedias can lead to interpreting them as being done by AI. When I was shown specific examples that raised concern, they didn't demosntrate AI use. However, every new interaction keeps bringing no answer to my repeated offers to solve the original Simple English Wikipedia issue and shift to new cases without any connection to the original topic which makes very difficult to find a solution. At this point, introducing new cases won't help solving the original problem.

To keep this discussion constructive and in line with normal dispute resolution practices, I will only focus on the Simple English Wikipedia matter. If i don't answer to further points, it is because they are not relted to that issue.

I ask again: What specific steps do I need to take to solve this issue and get my block reconsidered? Jon Gua (talk) 06:13, 26 March 2026 (UTC)

It might have originated from your simplewiki block, but that is not what this discussion is about (stewards rarely get involved with local blocks). If you missed Ferien's point in bold text ("you have used AI as an alternative to machine translations to create pages crosswiki where you don't know the language"), with the broader point having been mentioned here so many times, I don't know what else to say.

Regarding "To keep this discussion constructive and in line with normal dispute resolution practices, I will only focus on the Simple English Wikipedia matter. If i don't answer to further points, it is because they are not relted to that issue." – you are the last person in this thread to be in a position to determine how other people view or discuss a matter about your conduct across multiple wikis (and not just simplewiki). Putting it bluntly one final time without sounding like a broken record, this discussion is not about your simplewiki block, full stop. //shb (t • c) 06:27, 26 March 2026 (UTC)

I understand the points raised, but I am still blocked at the Simple English Wikipedia despite repeatedly providing evidence that I did not use AI there — my work consisted of summarising and rewriting articles from other Wikipedias, which is allowed. I have consistently offered to revise any articles that do not meet the Simple English Wikipedia guidelines regarding complexity, yet there has been no constructive dialogue addressing that.

Following what you said, now I understand that anything regarding the Simple English Wikipedia issue should be addressed on that project. Could you please confirm this? If so, I will focus on resolving the matter there, although I have not yet received a response regarding this topic, neither here nor on the Simple English Wikipedia.

Regarding other accusations of AI use outside the Simple English Wikipedia, I have explained every example and provided evidence to the best of my ability that my contributions are not AI-generated. I have tried to keep a polite and constructive tone at all times, focusing on solving issues rather than escalating the conflict.

I also want to clarify that examples that were cited as AI-generated (for instance, the Extremaduran article) were not creating using AI, as you acknowledged. I have consistently explained how my working method (summaring, adapting and referencig multiple Wikipedia articles) could seem unusual but it is not AI-generated. My aim throughout this conversation has been to keep it constructive and focused on verifiable issues.

What is more, I explained that most of my contributions across Wikipedias involve very little language-use, consisting mainly on adding categories, images, templates and interwikis. Furthermore, regarding the main accusation here (AI-usage across multiple Wikipedia) it has even acknowledged on this conversation that using AI is not prohibited, so even if that were the case (which is not as evidenced multiple times), you have even stated that it is not against the rules.

Finally, I want to point out that my goal is to solve the Simple English Wikipedia issue constructively so I will focus on solving that on that project. I acknowledge that some of finished articles could sometimes looked like done by AI, but I believe I have provided enough evidence and explanations for any concerns raised regarding AI use outside the Simple English Wikipedia and I have repeteadly describe my working methods on those Wikipedias, which I considere more than adequate. Jon Gua (talk) 08:20, 26 March 2026 (UTC)

What part of "this discussion is not about your simplewiki block" did you somehow miss in your slop of text? //shb (t • c) 09:46, 26 March 2026 (UTC)

Thank you for your clarification. As I mentioned before, I understand that this conversation is not about the Simple English Wikipedia block itself. Following your explanation, I will continue the discussion on the Simple English Wikipedia, which is the appropriate place to do so.

I would also like to conclude this conversation respectfully. I have noticed that some responses have come across as slightly dismissive, and I hope we can maintain a respectful tone here as my intention has always been to address the issues constructively. Jon Gua (talk) 09:53, 26 March 2026 (UTC)

Gone back to using AI to write your messages? //shb (t • c) 10:09, 26 March 2026 (UTC)

You are not addressing any of the points raised so there is nothing to answer. Without answering anything of what I say there is no meaningful conversation happening here. Jon Gua (talk) 10:49, 26 March 2026 (UTC)

The accuracy of the venetian (venetan) language on one of the pages in question is very poor. I can't say whether it was generated by AI or not. However, if there's reasonable suspicion of mass machine translations/AI-generated pages, I think deleting the page is appropriate.

The intent is not to discourage @Jon Gua from creating pages, but rather to suggest the production of less extensive and more precise contributions. Sinucep (talk) 14:14, 26 March 2026 (UTC)

I didn't use AI but if the quality is poor, please go for it a delete it. I will definitely make much less extensive contributions and focus on quality, for sure. Thank you for your comment. Jon Gua (talk) 16:53, 26 March 2026 (UTC)

Done All accounts blocked globally by Vermont. Count Count (talk) 16:43, 28 March 2026 (UTC)

OAuth permissions

Spam page on io.wikipedia

Status:    Done

Hello Stewards,

Please delete the following page on Ido Wikipedia (io.wikipedia). It is a Portuguese-language crypto/company PR spam page created on the Ido language wiki.

• Link: https://io.wikipedia.org/wiki/ZBXCX

Thank you. --~2026-17283-84 (talk) 18:46, 19 March 2026 (UTC)

Done please report such instances to GSR next time instead. A09|(pogovor) 18:48, 19 March 2026 (UTC)