Chatbots (and the large language models that power them) have become very popular; they can often output material much faster than humans can. Unfortunately, they cannot write as well as humans (yet), and their output is prone to hallucinations, false citations, and other errors. This has created a major cleanup burden at Wikipedia, as many editors (especially new ones) try their hand at using artificial intelligence to edit Wikipedia. You can help by identifying AI-written text, removing unsourced or inaccurate claims, and by identifying AI-generated images. For more information, see our AI Cleanup Guide.
The first iteration of the page was a draft containing user communication ("Key things to note before resubmitting", and then the AI incorrectly telling the user to "**Remove the decline/comment header lines.** In your current draft, the very top has these lines which mark it as declined:") and was subsequently speedily deleted under G15. Then Draft:Sanctuary (podcast) was re-created, declined at AfC, the AfC templates removed and then a very similar version pasted into article space.
The author has disclosed paid editing on behalf of Voyager Media, who produce the podcast that is the subject of this article. The article only cites WP:IMDB and a site called Techney.com, so SIGCOV is not likely to have been met. --Gurkubondinn (talk)09:59, 3 April 2026 (UTC)[reply]
Delete - with the COI issues we shouldn't keep in the current form and I'm not seeing sufficient RS to support even a one-line stub. JMWt (talk) 11:33, 3 April 2026 (UTC)[reply]
Appears to be a WP:LLMT of a page on fr.wiki. I tried engaging with the editor to explain that references needed to be verified before they could be added to en.wiki but all they did was to change some of the code to make it look like references had been accessed in 2026. This is not about the date but about human curation of references added to new pages on en.wiki and I have no confidence that the editor who started this page has understood the importance of WP:V and instead wants a shortcut to import content from fr.wiki without sufficient human review. JMWt (talk) 14:20, 2 April 2026 (UTC)[reply]
and I did not advise you to "change the access dates". I said that Each one needs to be checked for relevance and the details need to be verified by you or removed.JMWt (talk) 17:43, 2 April 2026 (UTC)[reply]
Further: Reference #2 on en.wiki is ref name="teleobs-2016-5-1" and appears in exactly the same place on fr.wiki as ref name="teleobs-2016-5-1". JMWt (talk) 19:02, 2 April 2026 (UTC)[reply]
Editor here. @JMWt, you’ve got it wrong here too. I speak French at a moderate level, but even though I’m still learning, I assure you this: I did not use any LLMs whatsoever. Plus, I was going to fix those dates too, but you just had to do this instead. That’s why I’m here now. Why are you doing this? Please, tell me now. Jibblesnark86 (talk) 17:42, 2 April 2026 (UTC)[reply]
The references all come from the same references on the French Wiki, which tell you the same thing. And as I said, I was going to change the access dates too to match this year. So what’s the problem here. Is getting the same references from another wiki a bad thing or something? Jibblesnark86 (talk) 17:48, 2 April 2026 (UTC)[reply]
they don't though. If you had taken the references from fr.wiki and added them to en.wiki they wouldn't have the same dates. You took the code from fr.wiki. JMWt (talk) 17:50, 2 April 2026 (UTC)[reply]
How else was I supposed to make the article then? Would you rather have me start from scratch than have me rewrite the code on enwiki? Is that what you’re saying. And anyway, are you a new page reviewer yourself. I’ve asked you this many times, never got an answer. Jibblesnark86 (talk) 17:54, 2 April 2026 (UTC)[reply]
Comment While it's fine to copy the citation template code when translating an article (and I wouldn't call it LLMT solely on that), I have doubts even a cursory check of the references was done given that long broken links were copied over as well... JumpytooTalk02:51, 3 April 2026 (UTC)[reply]
Comment Article is notable which is important, could be marked with a {{LLM-assisted translation}} template to be rewritten by an experienced translator. Jôhola07:40, 3 April 2026 (UTC)[reply]
This was first declined at AfC in Diff/1311901752 as bio (about a person not yet shown to meet notability guidelines) and ilc (a BLP that does not meet minimum inline citation requirements). After being re-submitted, a different user removes the AfC templates in Diff/1345720950 (with the edit summary theroadislong moved it) and then promptly moves the draft into article space in Diff/1345720974.
The iEyeNews website seems to exist (according to DuckDuckGo search results), but it does not load for me. If the website and article actually loads for someone else then this is obviously not a fabricated/hallucinated reference. --Gurkubondinn (talk)15:42, 2 April 2026 (UTC)[reply]
Delete as not notable. iEyeNews seems to have closed. I'd like to note that I was the AfC reviewer who declined the draft, and I find the attempt at misleading fellow editors to be contrary to the collaborative nature of our project. Kovcszaln6 (talk) 16:06, 2 April 2026 (UTC)[reply]
Article is obviously AI-generated and references do not match with content, such as reference five and the apparent services of the tower. elijahpepe@wikipedia (he/him)00:11, 2 April 2026 (UTC)[reply]
Delete - this is the third time just this week that an author came to defend his AI usage by using using AI to generate a response on the AFD discussion, these LLM-generated replies are getting out of hand. ロドリゲス恭子 (talk) 05:09, 2 April 2026 (UTC)[reply]
The following discussion has been closed. Please do not modify it.
:I'm the creator of this article and I can assure you this article is not AI generated. The proposed tower on the Gold Coast is worthy of an article on Wikipedia. Just read this
Neutrality: all articles need to be neutral -- not taking a position
Not original research: Wikipedia is not the place to report original results, or do original synthesis
Verifiable: everything should be sourced to a reputable source
Notable: the article should be about something or someone that is notable in the world, as determined by other sources
Free: don't copy other people's text. Everything you write on Wikipedia is released under a free license: other people can use it.
This article complies with all of these guidelines. The sources are very rialable and I reckon someone put this for deletion because of AI usage, which there is no evidence to claim it was AI generated. Before nominating an article for deletion. Read the article first as the article complies with all Wikipedia policies.
Delete Literally raw LLM output with barely any editing, and the above 'response' from the article creator above is even more of the same (telling us to 'just read this' and then more LLM output is not a good look). Please respond with only your own words, N. Cooksey. Nathannah • 📮00:44, 2 April 2026 (UTC)[reply]
Delete as non-notable promotion and AI slop. Mixing in some typos doesn't make you sound human, it makes you sound like an AI with poor spelling. ເສລີພາບ (talk) 06:29, 2 April 2026 (UTC)[reply]
The article was almost surely created through the use of LLMs. There is an ongoing cleanup thread dedicated to the recent edits of the author of this article on the AI cleanup Wikiproject. The user in the meanwhile has been indef blocked in the main namespace for the usage of LLMs and completely blocked some days ago for their daily evasions with TAs. I've already reverted his other edits, what is left are the articles created by him. There are multiple WP:AISIGNS: In the first version created by the author there were a lot of WP:AIFICTREF (broken external links); there are still multiple signs of WP:AILEGACY (for example revealing their close friendship and intellectual exchange or demonstrating a sophisticated understanding of geography and global economic dynamics) that are completely missing in the sources; and finally there are facts that are simply absent in the sources that should reference them, such as the fact that the geographic maps were produced together with Gastaldi or the fact that Ramusio's work influenced the works of de Bry and Purchas (both facts are actually true, but they come from this source and not from the one used in that part of the article, which is this one). According to Copyleaks it was 100% AI-made (whilst according to GPTZero it's a mix of AI- and human-generated text, but I remind that these tools are not always reliable). All in all, I think that's a pretty straightforward violation of WP:NEWLLM and that the best course of action is WP:TNT. --Friniate ✉ 14:32, 1 April 2026 (UTC)[reply]
Delete looks to be largely an llm translation from it.wiki including the same references "retrieved" long before the page existed on en.wiki. Incidentally I have noticed that this phenomena is becoming more common - editors add a new page with a large diff which includes references which have been lifted wholesale from another language wiki. Even if they are curating the content, at very least they are not carefully verifying the references. JMWt (talk) 14:48, 1 April 2026 (UTC)[reply]
Comment Would a !draft help? There are hints of notability for this book, but this version of the article isn't quite good enough for Wiki. Translating the It wiki article isn't impossible, just difficult, if LLM isn't used. I've used Gtranslate a paragraph at a time when translating things in the past and you have to rework several of the citations if they are used multiple times; it's a pain in the neck, but can be done. I'm not interested in translating it, but others might be. Oaktree b (talk) 18:49, 1 April 2026 (UTC)[reply]
I think it is likely that the topic is notable and would welcome an article written by a human. But that requires a complete rewrite with no unchecked content from an llm. JMWt (talk) 14:42, 2 April 2026 (UTC)[reply]
Just a note: the lead and part of the section about the structure were indeed translated from it.wiki (I noticed it now, thank you all for pointing it out), but the "content" section was entirely a product of the LLM (and the section about the structure is partly different from the original). --Friniate ✉ 15:30, 2 April 2026 (UTC)[reply]
Delete per nom and above, not seeing much on this book. Little in Google scholar as well. One presentation paper, which does not confer notability. ← Metallurgist (talk) 23:27, 2 April 2026 (UTC)[reply]
AI-generated article, and even if it wasn't, it's an overly-verbose paraphrasing of the blind taste test article, with some wine examples thrown in to puff up the reference list (along with the kind of pointless detail LLMs can't resist, like "blind tasting involves concealing the wine label or bottle shape...this can be achieved by covering bottles with plain paper bags or decanting the wine into neutral containers". And many references are hallucinated: Ref. 20's PMID resolves to a different paper, and Ref. 6 does not at all say what it's cited for. Textbook slop, let's delete it and be done with it. WeirdNAnnoyed (talk) 11:35, 31 March 2026 (UTC)[reply]
Keep The LLM issues can be resolved by reverting the edits that introduced the LLM edits (so restore this revision or a bit before that: [1]). JumpytooTalk01:59, 1 April 2026 (UTC)[reply]
Comment I don't get why voters are ignoring the old revisions, so I reverted the article to the revision I suggested in my vote. JumpytooTalk02:21, 3 April 2026 (UTC)[reply]
There's a WP:COI in that the editor of the magazine in question appears to be the person writing the WP page. Also they seem to be relying on an LLM although this would appear to be less of a problem given the above. JMWt (talk) 09:47, 31 March 2026 (UTC)[reply]
Delete: Some mentions here [2], that's not enough for notability. Most of the sources mention this web magazine, but aren't very detailed. I don't see enough coverage. Oaktree b (talk) 13:24, 31 March 2026 (UTC)[reply]
What constitutes sufficient coverage for notability? There’s a mention of it in a popular book and documentary of the time and multiple websites and archives. ~2026-20069-11 (talk) 05:09, 1 April 2026 (UTC)[reply]
COI issues mean that one almost inevitably thinks that a topic close to you should be a page on en.wiki
LLM issues include introducing errors, but if it is a topic that the person knows better than anyone else, almost by definition they are best placed to curate the LLM output. JMWt (talk) 08:20, 1 April 2026 (UTC)[reply]
Delete : No significant reliable coverage available for this publication on google search. Only available source is the news of their apology on various news media. Other than this, no other coverage. Rht bd (talk) 12:29, 31 March 2026 (UTC)[reply]
Salamun Alaikum
I updated the article with some high quality news sources like Daily Star, Prothom Alo and TBS. This news shows that the magazine is notable. Also I added a criticism section for making the article neutral. In future, it will be more updated. Please check the new sources and I request to keep this article.
Thanks but all four sources you added have almost same wording, basically saying the same thing and are based on a press release (their apology) by Chhatra Sangbad. That's not WP:SIGCOV. আফতাবুজ্জামান (talk) 19:34, 31 March 2026 (UTC)[reply]
She needs to do some work on the article, but she is not finished yet. This is a notable topic with important scientific content. She is still working on it. I don't think deletion is warranted -- this is a new editor learning the ropes on Wikipedia. Give her a chance to sort out the referencing link issues and tone please. A problem with the DOI link does not automatically mean that it was written by AI (although I understand why you would assume that in this age of ChapGPT). This article is part of a course project and I think she will be able to get the article to meet Wikipedia standards. SarahSalal (talk) 00:11, 31 March 2026 (UTC)[reply]
The editor is new to Wikipedia to the extent of writing completely unreadable and repetitive content and yet somehow has been able to work out how to reference using refname on the first attempt in a single large diff? That seems highly unlikely. JMWt (talk) 02:56, 31 March 2026 (UTC)[reply]
Delete That broken DOI for a paper that has a perfectly serviceable one is a bad sign. While Concordia University does not appear to have an AI policy overall, this could be an ethics violation depending on the rules set by the course instructor [3]. It is, at any rate, bad for the student to upload anything that is WP:G15-eligible like this. That's not the way to learn how to do research. Stepwise Continuous Dysfunction (talk) 23:51, 30 March 2026 (UTC)[reply]
This is not an LMM article. I am writing it myself. Something went wrong with the referencing format and the links don't work, but I am working on fixing it and all writing is my own. CanadianGreatLakes (talk) 18:15, 31 March 2026 (UTC)[reply]
Draftify or move back to user space. The best case is that this wasn't ready to be moved into article space. Among other issues, it contains language suited to an essay but not an encyclopedia article (e.g., relationships between Arctic amplification and mid-latitude weather should be continuously monitored). Stepwise Continuous Dysfunction (talk) 19:36, 31 March 2026 (UTC)[reply]
Delete highly likely that an LLM is in use because new editors do not suddenly learn perfect referencing code. Even beyond that, the content is not readable prose. This is not fixable without WP:TNT. JMWt (talk) 03:02, 31 March 2026 (UTC)[reply]
Thinking a bit more about this, a novice editor would have to
1. Know to use the reference creator wizard/tool
2. Know that they needed the DOI url to create the properly formatted code
3. Know when and how to use refname
4. Do all of that for multiple different references in a single diff
If the DOI itself is wrong, I think it is unlikely that the tool would work. So in order to get a correctly formatted reference code with the wrong DOI the editor would have to essentially be hand coding the references. And they'd have to be doing that for multiple references, with refnames, in a single diff without mistakes.
I am learning how to do this in Wikiedu. I have taken the courses that were assigned to me in this platform. I am still having trouble figuring out how to properly upload my references. I am working on it now and for the rest of the week. CanadianGreatLakes (talk) 17:58, 31 March 2026 (UTC)[reply]
It works when I found the paper on google scholar and added it using visualeditor in my sandbox. So somehow you would have to find a magic DOI that works in the visualeditor cite tool (which isn’t first result I got using google scholar) but doesn’t give a valid DOI url. It could happen but still seems unlikely JMWt (talk) 16:07, 31 March 2026 (UTC)[reply]
Maybe the editor was using their universities library search tool? That probably linked to the Wiley source which has the broken DOI. JumpytooTalk18:02, 31 March 2026 (UTC)[reply]
Draftify I think the issue with the doi has been adequately explained above and there is no harm if the page is preserved in some way. Kelob2678 (talk) 23:13, 31 March 2026 (UTC)[reply]
Per WP:NEWLLM there are a limited acceptable uses for llms, one of which is to assist with translation following WP:LLMT. This page was created with edits with fairly numbers of large characters which the editor listed as a translation of the Russian. The talk page shows that this was achieved with an llm and there is very little evidence that the translation has sufficient/any substantive human checking. JMWt (talk) 17:08, 29 March 2026 (UTC)[reply]
Misuse of LLM per WP:NEWLLM. Exhibit A is the words of the editor who created the page on the talk page: they have described using Claude to "verify the page numbers against the Harrison 2024 text". Clearly no effort has been made to verify this is actually correct, because that would take more time than doing proper referencing in the first place using human facilities to read the page numbers. JMWt (talk) 11:30, 29 March 2026 (UTC)[reply]
That would also run up against the narrative that AI can do anything and everything in seconds, and we puny humans with our "logic" and "fact checking" are troglodytes by comparison. As for the article, delete. It effectively has only one source, which clearly has not been vetted. If the creator cares enough, let them spend all the time they saved with AI fixing the article properly. If not, we don't need it anyway.WeirdNAnnoyed (talk) 12:47, 29 March 2026 (UTC)[reply]
I have removed all LLM-generated content and replaced it with a human-written article, about a book that I consider obviously notable (hundreds of scholarly articles about it). If you are willing to withdraw your delete !vote (since I believe your concerns have been addressed), this AfD can be speedy-closed. ~ le 🌸 valyn (talk) 18:40, 31 March 2026 (UTC)[reply]
Delete per nom. I checked the first two footnotes, which probably constitute a copyright violation. The article,
Les amours de Psyché et de Cupidon (The Loves of Psyche and Cupid) is a 1669 prose and verse work by the French author Jean de La Fontaine.[1] It is his longest single work, extending to nearly 140 pages in the standard Bibliothèque de la Pléiade edition.[1] The text is predominantly in prose but is framed as a prosimetric composition containing a number of inserted poems.[2]
The source,
Les Amours de Psyché et de Cupidon (‘The Loves of Psyche and Cupid’, hereafter LAPC) by Jean de La Fontaine (1621–95), best known for his Aesopic verse fables, is his longest single work, extending to nearly 140 pages in the standard Pléiade edition, and one of his few extensive compositions predominantly in prose, though its prosimetric frame contains a number of inserted poems.
Keep, and stubify to address copyright violations. This is obviously a notable book: with more than a thousand hits in google scholar, I immediately see dozens that satisfy WP:NBOOK. I am not persuaded AfD was the correct venue to address this problem, since the nomination does not address notability or give evidence of a WP:BEFORE. It could have been stubified as normal editing. Also, in the author's mild defence, it arguably did not contravene WP:NEWLLM at the time it was created, since at that time the guideline forbade creating articles "from scratch", and the Talk page comments do assert editing and human review in addition to the Claude review. Since copyright violations are urgent, I will stubify and revdel now. ~ le 🌸 valyn (talk) 22:27, 29 March 2026 (UTC)[reply]
That is obviously contradictory - the editor asserts human review.. but Claude introduced urgent copyright violation.
It's not enough and can never be enough because for every error that a careful and competent editor finds and corrects, the llm is likely the have introduced another 10.
Other editors need to go through this line by line and claim by claim, the llm has generated extra work. Yes, a human could have created mess like this but no they usually don't write content that looks superficially like normal WP coding and referencing in one large diff. We have enough problems already without the existential threat to this encyclopedia of machine written content. JMWt (talk) 06:01, 30 March 2026 (UTC)[reply]
That said, I can see that your stubifying has required some effort and I thank you for doing that. A discussion about a stub written by a human is a different question than a page mostly written by llm and if you are telling me that the page has indeed now been completely checked and references checked for relevance by you then I am happy to reconsider withdrawing the nom as it looks like the llm added content has been reviewed carefully by you. JMWt (talk) 07:14, 30 March 2026 (UTC)[reply]
All content currently in the article was fully human-written and verified by me. I have removed and revision-deleted all LLM-generated content because of the apparent copyvio. The prevalence of copyvio is why we changedWP:NEWLLM to now forbid this kind of use; I was just noting that this particular kind of article creation was permitted at the time when it was created, and the original author contributed in apparent good faith.
I would never argue that other editors are obliged to go through an LLM-generated article line by line and claim by claim. However, if an article does not qualify for WP:G15 speedy deletion (which is for LLM generated articles with no human review), I think it is reasonable to consider doing a WP:BEFORE and stubbing an LLM-generated article down to a single sentence as in this revision before bringing an article to AfD. I'm personally choosing to expand the article again after stubification because 18thC literature is an interest of mine, but a notable stub with a good infobox is a perfectly acceptable incremental improvement of the encyclopedia, and takes less total editor time than a full AfD. ~ le 🌸 valyn (talk) 19:55, 30 March 2026 (UTC)[reply]
I don't think TNT is necessary any more, since I have already removed all LLM generated content. The current state of the article is completely different and human-written by me. ~ le 🌸 valyn (talk) 19:56, 30 March 2026 (UTC)[reply]
The article was almost surely created through the use of LLMs. There is an ongoing cleanup thread dedicated to the recent edits of the author of this article on the AI cleanup Wikiproject. The user in the meanwhile has been indef blocked in the main namespace for the usage of LLMs and completely blocked some days ago for their daily evasions with TAs. I've already reverted his other edits, what is left are the articles created by him. There are multiple WP:AISIGNS such as WP:AILEGACY, but especially a lack of text-source consistency mixed with hallucinations: for example, in the text it's claimed that in Italy There are about 1 million hectares of olive groves in specialized use and another million in mixed farming. The claim is referenced with this source. The problem is that the source talks only about the first million, and there's no trace of the other million. Immediately above it's claimed that Central regions have 19% of the area [...] Only 2% is in the north. These informations are referenced with this source that doesn't talk at all of these percentages. They are true, but they come from a completely different source... Additionally, both GPTZero and Copyleaks confirm that the text is AI-generated. All in all, I think that's a pretty straightforward violation of WP:NEWLLM and that the best course of action is WP:TNT. --Friniate ✉ 14:25, 29 March 2026 (UTC)[reply]
Delete because llm slop is too difficult to cleanup. The editor who started the page has been blocked, with little other substantial edits from anyone else. The topic looks like it is likely notable. We just need actual humans to write it. JMWt (talk) 16:35, 29 March 2026 (UTC)[reply]
Ok, I see that you've completely rewritten the article, thank you. I'd withdraw the procedure except that I see some problems in the current article too:
In the source it isn't said that Italy suspended the production of olive oil during WWI. In fact, it'd be almost impossible I think. What's being said is simply that "France and Italy were otherwise occupied" and they didn't export as much olive oil as before.
In this source I can't find anywhere that the Late Empire still had at least 200 groves. There's a 200 olive oil production installation figure, but it's referred to the Late Republic.
I don't really understand why only the olive charcoals dating back to 5740–5590 BC where cited when in the source are listed also older archeological discoveries related to olive oil production...
I can't check other sources, but I find quite strange sentences like Italy and Spain together produce 96% of all olive oil that is exported from either Italy, Spain or Greece into the top fifteen importers of olive oil. What does it even mean?
I don't know if these are all remnants of the LLM-slop that there was before, but if we're going to withdraw the AfD procedure and keep the article, we need to make sure that every piece of information is really backed by sources. --Friniate ✉ 09:53, 30 March 2026 (UTC)[reply]
Good eye, #1–3 were issues with my own (human) reading and I had difficulty coming up with a none-CLOP way to reword the source for #4. I have fixed those to the best of my ability. Let me know if there are still issues. Rand Freeman(talk to me)10:04, 30 March 2026 (UTC)[reply]
Misuse of LLM per WP:NEWLLM. Exhibit A: a references claims "Retrieved May 18, 2024" when the page did not exist in 2024. The editor that created the page has detailed their misuse on the talk page JMWt (talk) 10:07, 29 March 2026 (UTC)[reply]
I think there is very little point in playing wack-a-mole, there are likely many errors one could find and many different reasons to !delete, the overwhelming one is use of LLM and it is not really worth the effort to find and argue about all the other details. JMWt (talk) 11:22, 29 March 2026 (UTC)[reply]
Delete, it's barely an argument with an unexplained and nonsensical error like that where it would even qualify for G15. Even if this were someone's old draft from 2024, the access-date should have been updated when the author (should have) re-read the source again before including it. --Gurkubondinn (talk)11:33, 29 March 2026 (UTC)[reply]
Delete: per G15. That access date error is a sure sign that the editor did not properly review the article, and admitting that it was created by AI is a headshaker. Ravenswing 15:35, 29 March 2026 (UTC)[reply]
G15 Delete In addition to above, the " History of the Clay-working Industry in the United States" links to completely unrelated book. I tagged the article for speedy. JumpytooTalk17:57, 29 March 2026 (UTC)[reply]
I agree that any weak, connected, or promotional sources should be discounted or removed, and any wording that appears non-encyclopedic should be cleaned up. However, I believe the subject is still covered by multiple independent sources that support notability, including Engineering.com (2021 and 2023), LSM (2020), Dienas Bizness (2020), and Tech.eu (2022). In my view, the key question is whether the topic has significant independent coverage, and I believe those sources are the relevant basis for evaluating that. Any prose or sourcing cleanup issues can be addressed through normal editing rather than deletion. Martin-Palmet (talk) 19:21, 28 March 2026 (UTC)[reply]
Article smells LLM generated and was created with HTML markup. All sources are either unreliable (forums/Reddit) or primary from Canon. Laura240406 (talk) 10:44, 28 March 2026 (UTC)[reply]
also have a look at all the other pages (starting from November 2025) created by the creator of the page, all follow this pattern and some have been struck down as LLM generated already Laura240406 (talk) 10:45, 28 March 2026 (UTC)[reply]
Delete under WP:NEWLLM and WP:GNG, with full prejudice. This page doesn't even properly document the protocol, and the sourcing is not even close to WP:RS. If the protocol should be documented on Wikipedia then it should've been done in a section of Canon RF lens mount, which this page doesn't even link to. There's no wikilinks at all on the page, which is another indicator of very sloppy chatbot use. --Gurkubondinn (talk)11:38, 29 March 2026 (UTC)[reply]
Comment: the artile was deleted under G15 by rsjaffe. I don't want to close the discussion because I !voted in it, and it's also not recommended for non-admins to close if the outcome was a deletion. --Gurkubondinn (talk)11:18, 30 March 2026 (UTC)[reply]
Speedy keep. I am not aware of any issues with the article. WP:LLMDISCLOSE cannot be grounds for AfD nomination, IMHO. Just in case, I have run verification on the claims in the article and a check for close paraphrasing that did not point to any major issues either (results are on the talk page). Викидим (talk) 20:24, 27 March 2026 (UTC)[reply]
Question: have you actually read all of the sources you reference in the article? Are you able to reference academic papers correctly? Why exactly have you used an llm - and do you have the skills to assess and critique the responses it gives? JMWt (talk) 22:03, 27 March 2026 (UTC)[reply]
Yes, of course, with a caveat. Article, contrary to the assumption by the nominator, is not entirely created from the LLM output. In addition to a lot of my manual edits, it is based on a translation from the Russian Wikipedia (ru:Смесоипостасная Троица), all of this BTW is clearly listed in the LLMDISCLOSE. Russian source proved hard to get, so I did not read it and simply trusted the author of the Russian article, see the details (yellow question mark) in the verification table on the talk page of the article. Викидим (talk) 23:00, 27 March 2026 (UTC)[reply]
You possibly misunderstood what I have just said: with the exception of one (!) 1924 source that is only used for one (!) trivial claim, I have read and checked the sources against all the article text. If the one short sentence that still requires the 1924 academic source for verification is a problem (it is A Synod decree from June 11, 1764, had already previously ordered that "strange and absurd indecencies" in iconography—specifically citing the three-faced, four-eyed Trinity presented to the Empress as resembling "Hellenic gods"—be suppressed), then surely this sentence can be deleted as opposed to the whole article? Викидим (talk) 06:10, 28 March 2026 (UTC)[reply]
We are both correct. Seeing where this train is going, I have fetched one Russian source and reviewed it, thus the discrepancy. The 1924 source is not marginal, just old and not widely available. To close the issue, I have spent some time and got the book online in the Boris Yeltsin Presidential Library. Will read it and update verification report in a short time. Викидим (talk) 06:55, 28 March 2026 (UTC)[reply]
Comment I am somewhat puzzled by this AfD. The article is pretty robust, with sources, sfn templates with page numbers, and now even a verification report on the talk page with quotes from sources. What percentage of our articles are at this level of development, even the new ones? For a comparison, let's look at the Miss Venezuela 1996 (we recently had a whole series of these restored to the article space): just two sources, each one used to confirm just one claim (some in the series have just one source), no traces of any attempt of checking the WP:V. I fully expect each of these articles with practically all claims unsourced to have multiple factual errors, yet they seem to generate no stir. On a common-sense level, concentrating on the "Trifacial Trinity" seems like wasting time. The purpose of the AfD IMHO is to weed out the bad articles, not the ones written with the help of a particular tool. Викидим (talk) 06:25, 28 March 2026 (UTC)[reply]
Keep – The editor who authored this article is competent at producing content for Wikipedia. He applied AI to assist him in creating the article, and then spent the effort required to make sure the article is worthy of keeping. WP:NEWLLM is not a content guideline, it is a behavioral one, because most people cannot be trusted to use LLMs responsibly. What do we do when an article is fixable? We fix it. This article has not only been scrutinized to determine that there isn't anything glaringly wrong with it, it is far past the threshold that if anything is wrong with it, it would probably be minor and definitely worth fixing. Therefore, we should keep the article. You have at least 2 experienced editors willing to further improve the article as needed (myself included), so look it over carefully, and point out any problems you can find with it, and they'll be fixed in short order. Sincerely, — The Transhumanist06:38, 28 March 2026 (UTC)[reply]
Which focuses on how it was edited (with what tools), that is, using AI, which is a behavior. Check the arguments. They are about the probability of screw ups with a tool in the hands of someone who doesn't know what they are doing. So, it doesn't apply here. AfD is about content, so let's focus on that. The content of this article has been gone over carefully to identify and remove any such problems. So, its development has already transcended those concerns and moved past that phase of development. It's in the tidying up stage now. Again, if there is anything wrong with the article, point it out, and it will be fixed. That's a major purpose of AfD: to determine what needs to be done to an article to make it worthy of keeping. — The Transhumanist07:59, 28 March 2026 (UTC)[reply]
I don't agree. The point is that if someone uses llms - to the extent of trusting the output more than themselves - then nobody can check it. Not them, not anyone else.
As I have alluded to above, the llm was asked to create a reference. The reference is wrong. The editor has not noticed and has not made any effort to correct it - and we are now to take it on trust that all the other things they have added, which I cannot quickly check, have been done carefully and to the best standards of en.wiki? JMWt (talk) 08:44, 28 March 2026 (UTC)[reply]
If you have discovered an issue (The reference is wrong), would you mind telling me what this issue is, so I can fix it? Викидим (talk) 08:55, 28 March 2026 (UTC)[reply]
no I can't. The whole point is that you seem unable to intelligently moderate the results of the llm. If you can't see what's wrong with the reference then maybe you need to have a bit of a think about how and why you are using llms like this. There's a perfectly usable system for creating references, so use it. Everyone else does.JMWt (talk) 09:11, 28 March 2026 (UTC)[reply]
Dear colleague, no I can't might mean different things: (1) "no, I do not know of any issue" (2) "yes, I do know about an issue, but cannot tell you". Which one is it? Ad hominem discussions If you can't see what's wrong with the reference then maybe you need to have a bit of a think ... do not improve the article. Викидим (talk) 09:32, 28 March 2026 (UTC)[reply]
we are not here discussing improving the article, we are discussing whether it meets the standards for inclusion and by your own words and deeds it does not. You have added slop to this encyclopedia and expect praise and thanks for doing it. JMWt (talk) 09:52, 28 March 2026 (UTC)[reply]
I could care less about the praise. As my profile in Russian says, "Through my work, I get regular, independent (often market-driven) proof that I'm right, so I don't need to stroke my ego by winning Wikipedia discussions just to keep my endorphin levels up." Here, I simply look for a straight answer to a simple question: do you know of any error in this article? Викидим (talk) 10:04, 28 March 2026 (UTC)[reply]
Delete - per WP:NEWLLM. The editor that did much of the work on the page has been open that they have used an LLM improperly and there is evidence that the LLM just isn't very good at writing Wikipedia articles. In future use your brain and ability and do not delegate to a statistical model. JMWt (talk) 08:47, 28 March 2026 (UTC)[reply]
(1) I did not say that I have used an LLM improperly. (2) When I was young, some very good linguists taught me that many theories of language declare it to be a statistical model (cf. Exemplar theory), and the goal of learning is to form or shift these distributions. AI at the time was just a way to get government grants, LLM did not exist. I therefore see no shame in using a statistical model working faster than my brain (similarly to the porter who does not deliver goods better than the delivery driver, just slower). Викидим (talk) 09:14, 28 March 2026 (UTC)[reply]
yes you did. You said how you used it which was clearly against WP:NEWLLM. There's absolutely should be shame in using LLMs to create obviously bad content. There's nothing else to be said, do not expect me to reply to you again. JMWt (talk) 09:18, 28 March 2026 (UTC)[reply]
Comment: There's a stark contrast in conduct here between the keep and delete sides that needs highlighting for the closer. Викидим has bent over backwards in good faith—disclosing AI use upfront with WP:LLMDISCLOSE, building from a ru:WP translation, manually verifying all claims (including sourcing a rare 1924 book mid-discussion via the Yeltsin Library), and providing comprehensive verification tables with quotes right on the talk page for anyone to check. That's about as in good faith as you can get.
Contrast that with the following deletion supporter's approach:
JMWt alleges a "wrong reference" but in the face of three direct requests to specify (so others can fix it!), responds "no I can't" and "do not expect me to reply again." That's not collaboration, it's disruption of the AfD process, that blocks improvement against the spirit of WP:BRD and WP:FIX.
And then there are the aspersions. There's no call for personal barbs here, and yet the attacks have piled up, right on this page: "Unable to intelligently moderate," "have a bit of a think about how you're using LLMs," "added slop to this encyclopedia and expect praise."
Slop? From an editor who discloses his methods, verifies his sources page by page, tracks down rare volumes from the Yeltsin Library while under fire, and lays out quotes for everybody to see?
its not an aspersion. You also are capable of looking at the references and determining if they meet the standard. Instead you are here making unverifiable claims about the competence of an editor using LLMs - who has by their words and actions by definition shown that they are unable to read, internalise and understand WP:NEWLLM to the extent that they cannot even understand the problem. There are only a small number of references on the page, so go back there, look critically at the references and have a think about why it is unacceptable. We don't use LLMs like that. End of. JMWt (talk) 19:12, 28 March 2026 (UTC)[reply]
Dear colleague, it looks like I have finally solved your riddle. You most likely refer to the cited work of Judova. If yes, there is no error there - and no AI involvement. This is apparently an unpublished work, so {{cite journal}} is not applicable. I would not have quoted it at all, but its importance was validated by a publication in a peer-reviewed journal (Thiessen 2018). Thiessen had quoted Judova in exactly this way - as a web reference to academia.edu. For precisely the reason of the missing publication place and date, this is the reference I have struggled with, and ended up handcrafting it. So whatever deficiencies are there, they are entirely my fault, not an AI running unchecked. If this is the issue you have refused to disclose, your arguments here do not have any foundation.
To avoid future confusions, I have now added a small note to explain the cite, the reason for its inclusion and the cite format. I have also added "n.d." to explicitly specify that neither I, nor Thiessen know the date the work was written on. Викидим (talk) 20:58, 28 March 2026 (UTC)[reply]
The first line of this god-awful essay literally says "Discuss the work of art in relation to medieval ideas about physiognomy and/or
Delete per NEWLLM and what JMWt is saying. It is the same argument that I have tried making countless times at this point, just presented much better thah I have managed to. The LLM has evidently added at least one false claim, and we are then being asked to trust the rest? In an article where those of us that don't speak Russian are unable to verify the sources ourselves? That is a gamble to slowly degrade the encyclopedia. --Gurkubondinn (talk)10:19, 28 March 2026 (UTC)[reply]
No, I am not going to stuff AI-generated WP:BEANS up my nose. You have the same capability as JMWt to find this reference, and if it is pointed out then you would just fix/remove it and claim that the problem is now fixed. But this is only a symptom, and it is a symptom that tells us that the sources have not been adequently read by the author, which shows that we cannot be asked to trust the rest of them either. --Gurkubondinn (talk)19:07, 28 March 2026 (UTC)[reply]
Exactly this. The fact that nobody has gone back and edited the reference to meet the normal standards for referencing on en.wiki and instead want to come and fight for the right to be respected whilst poorly curating the results of an llm says all that needs to be said. Can't Even See The Problem. JMWt (talk) 19:17, 28 March 2026 (UTC)[reply]
Academia.edu is not a publisher, it is a repository. We never reference anything without a date. That's two mistakes which an experienced editor who actually has access to the source they are citing would not make. As stated above, it is a symptom. A symptom of a poorly referenced article where the llm has written a nonsense reference that an editor with experience of writing and referencing academic articles would not make. The norms are very clear and obvious at WP:CITE. JMWt (talk) 19:39, 28 March 2026 (UTC)[reply]
on further investigation - which to be clear, I absolutely did not need to make - the source itself is unreliable and undated. It is likely either a university essay written by an undergraduate or is itself a product of an AI halucination. Either way, entirely unsuitable as a source on en.wiki. That's without considering the content, which is likely to be complete rubbish. JMWt (talk) 19:59, 28 March 2026 (UTC)[reply]
I have replied elsewhere on this thread. If you are still unhappy (Thiessen 2018 was OK with that), this reference can be removed. Citing undated work is perfectly normal, out cite templates even support the n.d. format for it, it is used countless times throughout Wikipedia. Викидим (talk) 21:10, 28 March 2026 (UTC)[reply]
Delete LLM trash. Competent human beings are required to write encyclopedia articles to avoid fabrication and plagiarism. The article says "By the 12th century, direct representations of the Trinity began to flourish ..." and the cited source says the exact same thing minus one word: "By the 12th century, however, direct representations of the Trinity began to flourish ..." Don't waste any more of our time with this type of thing. Asparagusstar (talk) 22:04, 28 March 2026 (UTC)[reply]
Delete While I understand the author tried their best to wrangle the LLM to be policy compliant, there are still some issues, such as here where the LLM picked the incorrect painting than what was used in the source. I also asked LLM to do source verification, and it found issues, with the most concerning one being a continuity error (A Synod decree from 11 June 1764 cites the "three-faced, four-eyed Trinity presented to the Empress", which earlier in the article says it given in 1767?). Combined with the other issues called out above, I don't have the confidence here to IAR NEWLLM in this case. LLMTRANSLATE cannot be used here, as per the prompt history provided by the author the base of the article was generated by the LLM, and the Russian WP article was only provided later. JumpytooTalk01:51, 29 March 2026 (UTC)[reply]
Delete, fails GNG. The author has admitted to using LLMs for creating articles and writing new content. And short as it is, this stub reads like it was generated, the statement that [it] was a popular format seems like WP:OR. --Gurkubondinn (talk)10:26, 27 March 2026 (UTC)[reply]
Comment - from personal experience, this was a somewhat common format for Macintosh software documentation in the early to mid-1990s; it was used by Apple themselves for the CD-ROM edition of Inside Macintosh, for example. Online coverage is going to be sparse given the time period; a comprehensive search of print sources might turn up more coverage. Omphalographer (talk) 20:16, 27 March 2026 (UTC)[reply]
Weak Keep Some sources (MacWelt, MacFormat) show the article subject may be notable. Sourcing could be better, but I expect other similar sources as presented here exist offline. Pavlor (talk) 05:18, 30 March 2026 (UTC)[reply]
The article was almost surely created through the use of LLMs. There is an ongoing cleanup thread dedicated to the recent edits of the author of this article on the AI cleanup Wikiproject. The user in the meanwhile has been indef blocked in the main namespace for the usage of LLMs. I've already reverted his other edits, what is left are the articles created by him. There are multiple WP:AISIGNS such as broken links that never existed and most of all clear hallucinations: for example, in the article it's being said that He participated in the battle of Tabasco, where he played a prominent role as royal factor,. This clearly doesn't make any sense, since a factor is a civil role, not a military one. In the cited source it's written Así, en Tabasco, fue designado factor real (u oficial encargado de recolectar los beneficios e impuestos del rey). [transl. In Tabasco he was nominated royal factor (and officer in charge of the collection of royal taxes)]. Why it's mentioned a battle then? Well, that probably comes from the First Battle of Tabasco, which was fought during the Mexican-American war, three centuries after Vázquez de Tapia's death. I think that's a pretty straightforward violation of WP:NEWLLM and that the best course of action is WP:TNT.--Friniate ✉ 12:24, 27 March 2026 (UTC)[reply]
Delete as a pretty clear NEWLLM violation. Disclaimer: I am an active participant in the WP:AIC project, though I have not previously been involved in the cleanup of this user. --Gurkubondinn (talk)14:30, 27 March 2026 (UTC)[reply]
Comment: The subject has enough sources to meet notability. But I'm unclear of the criteria for LLM generated articles. I know they're banned. But should they be draftified or deleted? — Itzcuauhtli11 (talk) 03:04, 30 March 2026 (UTC)[reply]
On the matter of sources, his book has been reviewed by notable scholars, there are contemporary sources about him, and other scholarly articles listed in es.WP as well. — Itzcuauhtli11 (talk) 03:09, 30 March 2026 (UTC)[reply]
@Itzcuauhtli11 I agree that the subject is notable. But fixing an article written by an LLM is actually more difficult and it takes more time than rewriting it from scratch. --Friniate ✉ 09:29, 30 March 2026 (UTC)[reply]
There are many issues with this page, the most obvious of which is that it's entirely AI slop. The original form of the article was intended to promote an AI website called aeronauticsmagazine.com, but that has since been removed. The user contesting the PROD asserted that this is a WP:SIA, but I don't see how it meets any of the criteria to exist as one: it isn't navigationally useful as none of the entries are bluelinked, there's no evidence that the topic of "magazines called Aeronautics" is notable, and the third criterion ("Short, complete list") states Lists in which no entry is notable are rarely appropriate. Even if the topic is notable and encyclopedic, WP:TNT applies, as there are no inline citations, making it unreasonably difficult to determine where the AI has hallucinated. lp0 on fire()09:26, 21 March 2026 (UTC)[reply]
I still think it meets either the first or third criterion for a WP:SETINDEX. The essay WP:TNT applies ONLY to articles "hopelessly irreparable" (first sentence). So keep and prune. Will give it a radical try. ~2026-16195-13 (talk) 09:59, 21 March 2026 (UTC)[reply]
Delete: no indication that any of the individual magazines are notable (if they were, we would create individual articles for them), and definitely no indication that the topic "magazines called Aeronautics" is notable. Rosbif73 (talk) 15:06, 31 March 2026 (UTC)[reply]
Consensus that "it is within admins' and closers' discretion to discount, strike, or collapse obvious use of generative LLMs" (Now in guideline: WP:AITALK)
"Most images wholly generated by AI should not be used." "Obvious exceptions include articles about AI, and articles about notable AI-generated images. The community objects particularly strongly to AI-generated images (1) of named people, and (2) in technical or scientific subjects such as anatomy and chemistry." (Now in policy: WP:AIIMAGES)
The WMF announced that machine-generated summaries of articles would be presented to readers, but then put the project on hold in response to negative community feedback.
AI Source Verification - Userscript that uses open-source models (Free!), Claude, Gemini (Free!) or ChatGPT to help check if a source supports a claim.
CitationVerification - Python script that uses MiniCheck and Claude to check if a source supports a claim.
South Korea enacts a law that requires human oversight of high-impact artificial intelligence (AI) systems, mandates disclosure and labeling of certain AI uses, and institutes fines of up to ₩30 million (around US$20,400) for violations. (Reuters)
