This is an advice page from WikiProject AI Cleanup. This page is not a Wikipedia policy, as it has not been reviewed by the community.

Shortcuts WP:AITALKSIGNSWP:AITALKSIGNS WP:LLMTALKSIGNSWP:LLMTALKSIGNS

This is a list of writing and formatting conventions typically found in comments written by Wikipedia users who are (or have been) suspected of using large language models or similar technology to write content, with real examples taken from talk pages and other discussion pages. Comments suspected of having been pasted from an LLM may be collapsed via {{collapse AI}} per WP:AITALK. Using AI to write comments is considered disruptive, and editors who do so may be blocked if they persist after being warned.

Although several of the tells mentioned at Wikipedia:Signs of AI writing (including boldface, em dashes, curly apostrophes and quotation marks, negative parallelisms, vertical lists, Markdown, and the rule of three) often appear in such comments, this project page only includes tells that typically do not appear in content added to articles or drafts.

Moreover, this list is descriptive, not prescriptive; it consists of observations, not rules. Advice about formatting or conduct can be found in the talk page guidelines, but does not belong on this page.

Most people on Wikipedia want others to believe that they're here for the right reasons and are willing to follow the rules, and may insist that they make their contributions with such rules in mind. AI chatbots, however, have a strong tendency to communicate this in a specific way: invoking policies, guidelines, and standards as a broad, formal, and abstract whole, as in legalese, and often using "AI vocabulary" to do so.

Examples

Thank you for raising your concerns. I understand that my recent rate of article creation has caused concern, and I apologise for not responding earlier.

My intention has been to improve Wikipedia's coverage of Malaysian state constituencies, which I understood to be comparable to other sub-national legislative constituencies that already have standalone articles. [...] My goal was to expand coverage consistently, not to create unnecessary work for New Page Patrol, AfC, or other editors.

Regarding the ongoing AfD discussion, I respect the community process and will not contest it by continuing to recreate or resubmit articles that are under discussion. My removal of the AfD template was a mistake on my part, and I now understand that it should not have been removed.

[...]

If the preferred process is to seek consensus first, improve a smaller sample of articles, or work in draft space before creating additional articles, I am happy to follow that process. I am also willing to improve the existing articles by adding more independent sources, expanding the prose, and addressing the issues raised during the AfD discussions.

My intention has always been to contribute constructively and improve Wikipedia's coverage of Malaysian elections and constituencies. I appreciate the feedback from everyone involved and will work within the community's expectations going forward.

— From this July 2026 ANI comment

I understand why I was blocked, and I accept responsibility for my actions. I used a large language model (LLM) to assist with my Wikipedia editing. At the time, I underestimated how incompatible that was with Wikipedia's expectations regarding editor responsibility, accuracy, sourcing, and the need for contributors to personally verify and understand every edit they publish. Regardless of my intentions, using AI in that manner was a mistake, and I recognize that I alone am responsible for the edits made under my account. Since being blocked, I have taken time to reflect on what went wrong and to better understand Wikipedia's policies and community standards. I now appreciate that AI-generated content can introduce factual errors, unsourced claims, and other issues that place an unnecessary burden on other editors. I understand that every edit must be based on reliable sources, carefully reviewed, and represent my own judgment as an editor. If I am given another opportunity to contribute, I will not use LLMs or other generative AI tools to write, draft, or expand Wikipedia content. I will make all edits myself, ensure that every contribution is supported by reliable sources with appropriate inline citations, and seek guidance through talk pages whenever I am uncertain. Rebuilding trust will take time, and I am prepared to do that through careful, policy-compliant editing. I sincerely apologize for my poor judgment and for the additional work my actions may have caused other volunteers. I respectfully ask that you consider giving me a second chance. If unblocked, I intend to demonstrate through my future conduct that I have learned from this experience and can contribute constructively while fully adhering to Wikipedia's policies and expectations. Thank you for taking the time to consider my request.

— From this July 2026 unblock request at the administrators' noticeboard

My intention was not to promote the subject but to document the career of a scholar whose achievements are supported by reliable published sources.

[...]

The article does not rely solely on institutional webpages or self-published material. It includes independent publications devoted specifically to Dr. Dmitrenco and her scholarly contribution. These include biographical essays, scholarly reviews, and bibliographical studies rather than routine mentions.

I fully understand that non-English sources may require additional effort to evaluate. If helpful, I would gladly provide English summaries or translations of the relevant passages from the cited publications.

I am an elderly contributor and not a native speaker of English. I have tried to follow Wikipedia's policies as carefully as possible.

— From this August 2026 AfC help desk comment

I would like to clarify that my intention is not to use Wikipedia for promotional or advertising purposes. The article was created to provide factual and encyclopedic information about me, my professional background, work, achievements, and contributions.

I understand that Wikipedia has strict guidelines regarding promotional content, notability, reliable sources, neutrality, and conflict of interest. I fully respect these policies and am willing to make any necessary changes to ensure that the article follows Wikipedia's standards.

My objective is not to promote any business, brand, service, or commercial activity through Wikipedia. I only wish to have an appropriately written, neutral, and well-sourced article about my professional journey and publicly documented work, provided that it meets Wikipedia's notability requirements.

— From this August 2026 AfC help desk comment

Links to searches

• "I understand the importance of adhering to Wikipedia's"
• "I fully respect Wikipedia's"
• "Moving forward, I will"^[a]
  • "Moving forward, I will adhere"
  • "Moving forward, I will ensure that"
  • "Moving forward, I will strictly"

Whereas AI-generated articles emphasizes a subject's impact, their creators' comments may exaggerate the extent to which a topic meets Wikipedia's notability guidelines.

Examples

The Pakistan Pharmacist Federation (PFF) is a well-established national organization in Pakistan, representing a significant and influential body of professionals in the pharmaceutical sector. The Federation plays an essential role in advocating for the pharmacy profession, advancing pharmaceutical education, and improving public health through various initiatives. Given its established presence, contributions to policy-making, and direct impact on the healthcare system, PFF qualifies as a notable entity under Wikipedia’s notability criteria for organizations.

[...]

Given the Pakistan Pharmacist Federation’s notability as a national organization with a significant role in public health and professional advocacy, and my commitment to revising the article to ensure it complies with Wikipedia’s editorial standards, I respectfully request that the speedy deletion nomination be reconsidered. I am confident that once revised, the article will meet Wikipedia’s requirements and provide valuable, neutral information about the Federation’s role in Pakistan’s pharmaceutical sector.

— From this May 2025 comment at User talk:Tahanazir

Over the past several months, I have worked to address the concerns raised during the review process. The draft includes significant coverage from reliable, independent sources that are not connected to the subject. Based on Wikipedia's notability guidelines, I believe the available sourcing demonstrates that the subject meets the criteria for inclusion.

Lovern is a notable figure within the comic book industry, particularly for his pioneering role in bringing computer colouring to comics through his company. Given the scope of his contributions and the available independent coverage, I am finding it difficult to understand why the article continues to be declined.

— From this June 2026 AfC help desk comment

When called out for using an AI to write or rewrite their content, some users will admit to using a large language model or similar technology to write or rewrite their content, but then follow that up with an assurance that they have checked their content to ensure it aligns with Wikipedia's editorial guidelines, or that the words in their comments truly reflect their thoughts.

Examples

I have reviewed WP:AITALK, and I am fully aware of the guidance. My comments here represent my own editorial judgment and responsibility. Suggesting otherwise is not appropriate.

— From this February 2026 comment at User talk:Grayfell

Fair one, I'm not trying to dismiss this. I am an intermediate English speaker. I used Grammarly’s editing tools (including grammar checking and clarity suggestions) to ensure my formal AfD nomination followed encyclopedic prose standards. Per WP:AI, using AI assistance for language and formatting is permitted provided factual claims are verified by the editor. I manually verified every source and citation in my nomination as evidenced by my correction of the URL error when I discovered it during source-checking. The substance of the nomination stands regardless of the drafting assistance used.

— From this June 2026 comment at Wikipedia:Articles for deletion/Rockstar (2026 film)

On the AI question? yes, I’ve used AI tools, but in a specific and within the policies way. I’m Spanish spoken, English isn’t my first language, and I’ve used AI to help with grammar, phrasing, and to check the structure of longer pieces. I haven’t used it to generate the actual content. Every fact, every citation, I’ve gone back to the source material myself and verified it before publishing.

---

Also, just to clarify how I work and why the text flagged, I do all my actual research and writing in Spanish first. Since English isn't my native language, I just run my finished text through a basic translator to pull it over into English. The ideas are 100% mine, but translation software naturally uses very predictable phrasing patterns, which is exactly why those AI detectors flagged it.

---

So to clarify the talk page posts, as well my process, is that I think and write out my thoughts in Spanish first. Since English isn't my native language, I can make some grammar errors, so I run my finished text through a basic translator to post it here. The actual thoughts and questions, are mine, but because standard translation software uses predictable phrasing patterns, it can end up looking like AI generated text to detector tools.

---

I want to explain how I work to answer some of the questions about automated text. I am not using a Large Language Model (LLM) to write my posts or generate content. My native language is Spanish, which is the language I use to do my research, map out my thoughts, and write my initial drafts. Because English isn't my native language, I just use a basic translation tool to turn my own finished writing into English so I can share it here. The predictable phrasing patterns that naturally happen with standard translation tools are exactly what triggered the automated AI detectors. [...] The research, ideas, and facts in my edits are entirely the result of manual human effort.

---

I am nearly fluent in English, but as a technical writer, I choose to use standard drafting and translation tools specifically to ensure that the structure, clarity, and paragraph flow meet the right standard for the reader.

Using tools to help organize ideas, check grammar, or improve structural formatting is entirely permitted under WP:LLM. It is also VERY discriminatory to weaponize WP:CIR against an editor simply because English is their second language, especially when the underlying research, facts, and citations are completely verified by human effort.

---

I am not disregarding the guidelines, nor discussing in bad faith, I am applying them exactly as they are written. The facts, research, and citations in my article are 100% verified by human effort.

— From all of these July 2026 comments at the AI noticeboard

When a user sees that an article they added content to has an {{AI-generated}} tag on it, or a draft they have worked on is declined for being AI-generated, or they are accused of using AI to write their comments, they often act as though they have no clue as to what is wrong with their content. They may request additional input or assistance to help them understand exactly where they are going wrong and how they can improve. They want to know precisely which passages are causing concerns so they can make edits to address those concerns.

Expressions of willingness to take constructive criticism for edits are very common in AI comments by users whose drafts were declined by reviewers, or in rare cases, themselves. The difference between this sign and similar reassurances by conscientious humans will generally become obvious after you actually provide that criticism.

Examples

I understand GPT Zero flagged it, and I know those tools aren’t perfect, but I also understand why you have to take it seriously. If it helps, I’m happy to go through the article section by section and show where each claim comes from. I want to contribute to this community properly and I’d sincerely appreciate any guidance on how to make my editing process more transparent going forward

---

I’m not trying to argue in bad faith, I really want to understand where the line is so I can stay clear on the right side of it. If you can point me to a specific passage or edit that concerns you, I’m happy to address it

---

If there are factual errors, issues, or passages in the article that you think are problematic, please point them out so I can fix them. Just saying there are "negligible improvements" doesn't help me understand what you feel is actually wrong with the content or the sources. I'm open to constructive feedback on the text itself

---

I'm completely open to constructive feedback on the text itself, and I'm happy to manually polish the phrasing going forward so it reads as naturally as possible.

---

If there is a concrete sentence in the article that violates a core content policy like verifiability or neutral point of view, please point it out so I can address it ASAP.

— From all these July 2026 comments at the AI noticeboard

Hi Gurkubondinn, Could you please review a new version which I have uploaded, and which has been rejected solely due to your previous rejection. The article is entirely manually written and I believe it meets Wikipedia's standards. I would specifically like to know what in my writing resembles an AI-chatbot? Obviously it is easier to be critical and delete on vague grounds than to approve. I strongly believe this version deserves to be approved. Thank you for your consideration.

---

Hi Gurkubondinn, Thank you for reviewing the second version. Could you please point to specific sentences which make my submission an "AI-generated" version? I am flabbergasted. Thank you.

---

Hej Gurkubondinn, unfortunately that comment doesn't help much. I've visited the WP:WHERESTHEAI page to no avail. I'm not contesting the review, but instead try to learn how I should not sound like an AI bot when writing myself! Obviously, my mother tongue is English, and undoubtedly this may influence my way or writing (in English that is). If you help me understand which sentences in the second draft are "copied from the chatbot or AI service", I'd be more than happy to take a look at them.

---

But seriously, you still have not pointed at the specific sentences which resemble a chatbot or AI-generator. Could you please do that so my Wikipedia experience will be a rejuvenating and refreshing one, as opposed to a mere frustration where I leave the seemingly wonderful resource without learning anything?

---

Hi there, My second draft about Blue Music Group was rejected with the comment that it "still obviously contains AISIGNS (most notably WP:AIATTR)." I rewrote the draft from scratch using primarily an independent All About Jazz profile and removed most self-published sourcing. I'm not disputing the review; I'm genuinely trying to understand what specifically in the current text appears to violate WP:AIATTR. Could someone take a quick look and point to the passages that should be rewritten?

— From all these July 2026 comments at User talk:Nycmoomin

Links to searches

• "If there are specific concerns about" OR "If there are any specific concerns about" OR "If there are specific concerns regarding" OR "If there are any specific concerns regarding"
• "If there are specific areas that" OR "If there are any specific areas that" -"know if there are" -"and if there are"
• "If there are specific sections that" OR "If there are any specific sections that"

When confronted by criticism of their work, it is common for LLM editors to make a stock request for an "experienced," "uninvolved," or "neutral" editor or administrator to provide a second opinion of their writing.

Examples

Hi, it's been 2 years since the warning banner has been placed and the article has been improved significantly. Would an independent editor please review the article to see if the banner is still appropriate? If not, would someone please remove the banner? Thank you!

— From this May 2026 comment at Talk:European Science Foundation

I respectfully request that an administrator:

• Review the article and its sources.
• Provide guidance on what specific changes would be needed to address the concern.
• Restore the article and template if they are found to be valid contributions.

— From this June 2026 ANI comment

When a draft is declined, LLM users tend to request source assessments in a similar manner. This usually carries the (unintentional?) implication that the reviewer didn't do their job properly.

Could an experienced editor please advise which of these sources, if any, count as reliable, independent, significant coverage for a biography article?

— From this May 2026 AfC help desk comment

Could an uninvolved editor advise which of the current references, if any, count as significant independent secondary coverage for notability purposes, and which should be treated only as supporting or weak sources?

— From another May 2026 AfC help desk comment

Could an independent administrator assess these specific sources?

— From this July 2026 AfC help desk comment

Links to searches

• "Can an uninvolved editor" OR "Could an uninvolved editor" OR "Would an uninvolved editor"

AI chatbots cannot read the decline notice left on the draft, and can only respond to whatever information regarding the declined draft that the user gives them. As a result, AI-generated questions regarding declined drafts will often express uncertainty or request clarification over the reason a draft was declined, even if the decline notice on the draft is very clear.

Examples

Is the primary issue related to notability, sourcing, tone, conflict of interest, or article structure?

— From this May 2026 AfC help desk comment regarding now-unavailable revisions to User:Revathyradhakrishnan2412/sandbox, which contained AI-generated text that went into detail about WP:CORPROUTINE activities

Could you please provide more specific feedback on the reasons for the decline? In particular, I would like to understand: Whether the issue relates to notability, sourcing, or tone If any specific sections need revision or removal What changes would make the article suitable for acceptance I’m committed to aligning the article with Wikipedia’s standards and would greatly appreciate your guidance.

— From this April 2026 AfC help desk comment regarding Draft:Ismatzein, which was said to have been promotional and failed WP:NBOOK

I would like some feedback if the issue is tone, sourcing, potential close paraphrasing, or article structure.

— From this June 2026 AfC help desk comment regarding Draft:Ontario employment Equity Act, which was declined for appearing to contain LLM-generated text

Is the concern related to factual accuracy, source use, writing style, article structure, or another issue?

— From this June 2026 comment at User talk:Helpful Raccoon regarding Draft:Digital Capital, which was declined for appearing to contain LLM-generated text

They may even compose self-contradictory comments that simultaneously acknowledge the specific reason for the draft's decline, but go on to say that they are trying to understand whether that was really the reason, or whether it was declined for some completely unrelated reason.

I am requesting guidance on Draft:Alexis Marcou after repeated AfC declines... I am trying to understand whether the issue is mainly source quality, article tone, or both before making any further changes.

---

The latest decline says the draft needs multiple published secondary sources that provide significant coverage, are reliable, and are independent. I understand the concern and do not want to keep resubmitting without first understanding the source issue.

[...]

I am not asking for the draft to be accepted here. I am trying to understand whether the problem is mainly that these sources are not strong enough for WP:BIO / WP:NARTIST, or whether the draft structure and tone are still the main issue.

Any guidance on which sources should be kept, downgraded, or removed would be appreciated.

— From these May 2026 AfC help desk comments regarding Draft:Alexis Marcou, which was declined for failing WP:NPERSON and reading like an advertisement

The rejection reason given was that the subject did not meet Wikipedia's inclusion criteria. After I reached out to the reviewer for clarification, they explained that they believed the decline reason should have been related to verifiability/sources instead and referred me back to the previous reviewer comments.

I'm confused because I'm not sure which issue I'm actually supposed to address.

[...]

I'm trying to understand whether the main issue is insufficient independent coverage to establish notability, the way I've used primary sources, or another sourcing issue that I'm misunderstanding.

I'm not looking to dispute the reviewer's decision. I genuinely want to understand what policy I'm misunderstanding so I can learn from the feedback and improve my drafting in the future.

---

Just to clarify, is the main concern that claims about the subject need stronger sourcing/removal of unsupported statements, rather than the subject’s notability itself? I’m trying to understand the specific issue so I can improve the draft appropriately.

— From these July 2026 comments at the AfC help desk regarding Draft:Jesse Ridgway (McJuggerNuggets), which was rejected for not meeting Wikipedia's inclusion criteria

Links to searches

• "clarify whether the concern" OR "clarify whether the issue" OR "clarify whether the main" OR "clarify whether the primary"
• "understand whether the main" OR "understand whether the primary" OR "understand whether the concern" OR "understand whether the issue" OR "understand whether the problem"

When a user is confronted for allegedly pasting content from a large language model, they will likely dismiss these concerns as unsubstantiated speculation based on tone or writing style rather than hard evidence of LLM use. Most if not all users who do so also tell those who confront them to just point out what needs to be improved and focus on that instead of calling them out.

Examples

“Reeks of LLM” is pure speculation. You may believe that statement strengthens your argument, but in practice it weakens it. It reinforces my concern that the discussion is becoming increasingly speculative rather than evidence-based.

---

You have not provided conclusive supporting evidence. You have speculated repeatedly, and you continue to rely on speculation rather than demonstrating identifiable policy violations or factual inaccuracies.

[...]

Quite frankly, unsupported speculation and subjective interpretation can be more problematic than AI itself, because at least AI concerns can be evaluated through evidence, sourcing, and verification standards.

---

I'm glad you helped. I understand what the guideline says. My point is that much of this discussion has moved beyond policy interpretation and into subjective speculation about tone, writing style, and assumptions regarding intent.

If there are factual inaccuracies, fabricated citations, or identifiable policy violations, point to them directly. Otherwise, repeatedly asserting that something “sounds like AI” is still speculation regardless of who drafted the guideline.

The issue is not whether LLM-generated article content is prohibited. The issue is whether conclusions are being drawn based on evidence or based on personal impressions and emotional interpretation of writing style.

---

Until consensus is reached through proper discussion and policy-based reasoning, I will continue removing warnings and templates that I believe are unsupported, speculative, or improperly applied. That includes AI-generation notices as well as criticisms such as “written like a resume,” which, if anything, further supports my argument that much of this discussion has shifted toward subjective interpretation of writing style rather than identifiable policy violations or factual inaccuracies.

— From these May 2026 comments at the AI noticeboard

Absent specific issues, a general impression about how the text was produced is not actionable.

---

You've just confirmed that your entire basis for this accusation is a personal impression formed by reading text. That's precisely what WP:NOLLM states is insufficient grounds for sanctions: "evidence beyond basic stylistic or linguistic indications." You have no method. You have a hunch. That's not actionable, and it never was.

Just look at how much time you spent on this exchange. It's time that could've been used to improve an article or do something else productive. The conclusion is, you just read a few words and formed an impression. That's all this ever was.

---

And no, I'm not going to answer a question built on an unsubstantiated premise.

---

As for conduct, I have engaged with every substantive point raised in this exchange. What I have not done is answer a question built on an unsubstantiated premise, which I am under no obligation to do.

---

More importantly, you have spent this entire exchange failing to identify a single factual error, a single unsourced claim, or a single content failure in the edit in question. Everything else has been an impression, an accusation, and now a hollow warning. The record of this conversation speaks clearly for itself.

---

Identifying a stylistic pattern is not evidence of AI generation, it's the same subjective impression restated with a grammatical label.

— From all of these June 2026 comments at User talk:Galilean-moons

In many cases, editors have responded to accusations of AI use by suggesting that the question of how their content was made is not as important as the question of whether their content adheres to Wikipedia's other policies and guidelines. These statements often take the form of negative parallelisms such as "Let's keep this focused on X rather than Y", "Instead of X, this discussion has veered off into Y" or "the correct course of action is to X, not Y".

When told that the general pattern of their editing has been disruptive, they often dismiss the accusation as being merely speculative, unfounded, or having no basis in policy, and accuse their critics of acting in a manner that they believe is uncivil, hostile, dismissive, unprofessional, counterproductive, or an affront to Wikipedia's collaborative nature. Persistence from others is interpreted as aggression; if challenged sufficiently they may state their intent to withdraw from the discussion as it has become "emotionally charged". Since the editor did not write this comment themselves, they will often continue the discussion or return to it shortly afterwards.

AI will seldom confirm that it's being used without the prior consent of the user, so if the response is vague, sidesteps the question or considers it to be a personal attack, this may be a sign of AI use.

Examples

Highlighting key:

• "Instead of making accusations, let's focus on fixing this article."
• "Let's focus on fixing this article instead of making accusations."
• "I consider your accusations to be frustrating and inconvenient."
• "The correct course of action is to help me fix this article, not to make accusations."
• "Instead of helping me fix this article, you kept making accusations."
• "You have continued to make accusations instead of helping me fix this article."
• "The correct course of action is to help me fix this article. Making accusations is a bad idea."

The goal here should be evaluating the actual substance and accuracy of the content, rather than gatekeeping contributions based on an editor's language background, and your argument sounds very discriminatory to me. I remain open to constructive, policy-based feedback on the article itself.

I haven’t yet been provided with the REAL SPECIFICS of where my article is wrong.

---

I am not disregarding the guidelines, nor discussing in bad faith, I am applying them exactly as they are written. The facts, research, and citations in my article are 100% verified by human effort. If there is a concrete sentence in the article that violates a core content policy like verifiability or neutral point of view, please point it out so I can address it ASAP. Otherwise, blanket dismissals based purely on stylistic generalizations go directly against explicit community policy text.

---

The Request: Let's stop the personal remarks and keep this professional. If you have an actual issue with a specific source or fact in the article, point it out and I will fix it.

---

Rather than speculating about how the text “looks”, I’d appreciate concrete feedback on any statements you believe do not meet Wikipedia’s standards, so I can improve the article on that basis. This is about the fifth time I’m asking for this.

— From these July 2026 comments at the AI noticeboard

I am concerned that invoking WP:RGW to strip out verifiable Indigenous history risks creating a one-sided narrative and leaning into historical erasure, rather than maintaining neutral point of view (WP:NPOV) and proper due weight (WP:DUE). Furthermore, the discussion has involved aggressive tone and assumptions about my identity rather than focusing strictly on source text (WP:CIVIL). He assumed that because of the way I wrote, I must be an AI program. I am not. I am clearly human.

— From this August 2026 ANI report

When confronted about using AI to write content, some users falsely claim that Wikipedia allows large language models to be used to generate content, and that whether the content they generate adheres to Wikipedia's other content guidelines matters more than how the content has been produced.

Examples

Please remember, Wikipedia allows tools like AI as long as the content remains factually correct and properly cited.

---

AI-assisted drafting is not prohibited. Per Wikipedia policy, the focus is on verifiability, not the tool used for drafting. Every fact in this article is supported by high-quality, third-party sources. Claiming "AI" as a pretext to delete documented financial records is a violation of WP:POINT.

— From this October 2025 comment and this March 2026 comment at Talk:Arthur Katalayi

I can confirm that Grammarly is the tool I use for proofing. I also think it’s important that others participating in this discussion recognize that Wikipedia does not maintain a formal prohibition on the use of AI-assisted tools. What ultimately matters is whether the content complies with Wikipedia’s core policies regarding verifiability, neutrality, sourcing, and accuracy.

---

Wikipedia’s guidance regarding AI focuses primarily on preventing hallucinated or fabricated information. It does not prohibit the use of AI-assisted tools altogether, particularly when the content itself remains verifiable, properly sourced, and subject to human review and editing.

— From these May 2026 comments at the AI noticeboard

When AI chatbots try to cite Wikipedia's policies or guidelines by mentioning project page shortcuts, they have occasionally misattributed ficticious policies or guidelines to specific project pages, including to project pages that were never intended to be cited in such a manner to begin with.

Examples

WP:NOTENGLISH clearly states:

"Wikipedia articles should not contain material based solely on non-English-language sources without reliable English-language sources."

In this case, the material you’re suggesting must be accompanied by a reliable English translation or reliable secondary sources for verification. Without that, it cannot meet Wikipedia’s verifiability standards.

— From this April 2025 ANI comment

Per WP:LAW: "Wikipedia is subject to the laws of the countries in which it operates."

— From this August 2025 revision to User:Flanker235

• WP:BLOCKPOLICY — blocks are preventative, not punitive.
• WP:BEFORE — admins should attempt communication before blocking.
• WP:AGF — assume good faith.
• WP:BITE — do not bite newcomers.
• WP:INVOLVED — avoid actions where involved.
• WP:UNBLOCK — blocks may be overturned when not aligned with policy.

— From this December 2025 unblock request

In some cases, while citing a guideline, an AI chatbot may state information that can be found at a related project page, but will fail to mention that other page and instead misattribute what it says to the guideline.

  Correct for WP:LLM

  Correct for WP:LLMS § Writing articles

  Incorrect anywhere

I want to make sure I understand the policy correctly though, WP:LLM does list permitted uses beyond spelling and punctuation, including using AI to “spot structural problems,” “generate ideas,” and ask “how to improve a paragraph.” My use was in that advisory space, with all prose written and verified by me. I’m not trying to argue in bad faith, I really want to understand where the line is so I can stay clear on the right side of it.

---

According to the policy documentation at Wikipedia, Large language models (WP:LLM), my workflow is explicitly protected. Under the section "Using LLMs as an advisory tool", the policy exactly permits editors to use digital tools to "suggest refinements to their own writing, and to incorporate some of them after human review, provided the LLM doesn't introduce content of its own." It states that asking "how to improve a paragraph" is a permitted function. As well, under the "Enforcement" section of the policy, it states: "The writing style of some editors may appear similar to the output of LLMs, meaning that the imposition of sanctions requires evidence beyond basic stylistic or linguistic indications." The guidelines straightforwardly command reviewing editors that "When evaluating possible LLM use, it is best to consider the full pattern of the editor's recent edits and whether the edits comply with core content policies."

— From these June 2026 AI noticeboard comments

In some cases, users of AI chatbots have pasted text containing hallucinated shortcuts that do not redirect to any existing page at all.

Examples

WP:ANACHRONISM – The sovereign state of Pakistan was established in 1947. No political, cultural, or geographical entity called "Pakistan" existed before this date. Using this name for 5000 BC history is a textbook anachronism.

---

However, every argument I have made is based on Wikipedia policies (WP:SYNTH, WP:NPOV, WP:ANACHRONISM) and mainstream academic reality.

• Contradictory Logic: The editor claims my edit is skewed by personal affiliation (WP:COIBELIEF), yet relies on personal YouTube observations to dismiss standard ecclesiastical terminology. [...]
• Misuse of COI (WP:COIBELIEF): Religious background is explicitly not a conflict of interest, and adding documented internal self-designation does not violate WP:NPOV.

When LLM users get into disputes, they tend to cite Wikipedia's policies and guidelines in their arguments by mentioning them alongside their corresponding shortcuts, often like this:

verifiability (WP:V), neutral point of view (WP:NPOV), and reliable sources (WP:RS)

In some attempts at policy shopping, an LLM user might claim that a project page says something it doesn't or cite a shortcut that hasn't been created.

Examples

Moving forward, I will strictly adhere to Wikipedia’s Neutral Point of View (NPOV) and verifiability guidelines, ensuring that all contributions are non-promotional, well-sourced, and align with Wikipedia’s standards.

— From this February 2025 comment at User talk:Gopaddi

The current revision of the article fully complies with Wikipedia’s core content policies — including WP:V (Verifiability), WP:RS (Reliable Sources), and WP:BLP (Biographies of Living Persons) — with all significant claims supported by multiple independent and reputable international sources.

— From this October 2025 comment at Talk:Arthur Katalayi

Thank you for your question. Yes, I did use an AI tool to assist with parts of initial drafting process. However, I understand that Wikipedia requires all content to meet its core policies, including verifiability, neutrality, and no original research.

As part of the ongoing rewrite, I am reviewing and rewriting the article in my own language to ensure it fully complies with Wikipedia’s guidelines, with proper sourcing and independent verification. I appreciate your patience and any constructive suggestions on improving the article.

— From this April 2026 comment at Talk:Black Diamond Museum, Angul

Links to searches

• "policies on Neutral Point of View (WP:NPOV)"
• "policies on Verifiability (WP:V)"

Users confronted for suspected LLM use often use the word concrete as an adjective, especially whenever they insist on being given "concrete examples" of text that needs improvement or "concrete evidence" that they used AI asides from "stylistic indications". When they don't get the "concrete" answers they demand, they dismiss the AI accusations against them as subjective speculation.

Note that humans have often used the word concrete in this sense when discussing the distinction between abstract and concrete things.

Examples

**Review:** The use of "significantly more" is subjective and requires specific figures to validate this claim. Without concrete financial data from reliable sources, this statement could be misleading or exaggerated. The source provided is a blog. — From this August 2024 comment at Talk:Eric Dick (lawyer)

Where is the concrete evidence linking me to these subjects?

---

I want to make it clear that I have never been paid for any edits, and it’s disappointing that such assumptions are being made without concrete evidence.

— From these October 2024 comments at User talk:Damjana12

In the absence of concrete evidence, I propose removing the AI-generated tag immediately to maintain the article's integrity. — From this October 2025 comment at Talk:Slavery in Portugal

If you have concrete, actionable concerns about particular sections, please feel free to cite them directly.

---

In short: policy does not support tagging based on hypothetical, unsubstantiated, or historical speculation — particularly when no editor, including yourself, has identified any concrete violation after repeated invitations to do so.

---

Until any concrete issues are presented, the article remains compliant and accurate, and the LLM advisory banner is no longer warranted.

---

A generalized allegation of puffery, absent concrete examples or explicit policy-based objections, is insufficient.

— From these October 2025 comments and this December 2025 comment at Talk:Arthur Katalayi

Without concrete examples, your concern cannot be evaluated in line with WP:V, WP:RS and WP:BURDEN. — From this April 2026 revision to Talk:House of Dust (architecture)

I think you're the one who's incapable of being an editor if you can't express concrete things, but rather general concepts that have little meaning. — From this June 2026 comment at ANI

If there are concrete concerns, such as inaccuracies, sourcing problems, or wording that does not meet Wikipedia standards, please point them out so they can be addressed. — From this June 2026 comment at User talk:Galilean-moons

If there is a concrete sentence in the article that violates a core content policy like verifiability or neutral point of view, please point it out so I can address it ASAP.

---

I’d appreciate concrete feedback on any statements you believe do not meet Wikipedia’s standards, so I can improve the article on that basis.

— From these July 2026 comments at the AI noticeboard

Historically, some comments generated by AI chatbots have begun with text that appears intended to be pasted into the Subject field on an email form.

Examples

Subject: Request for Permission to Edit Wikipedia Article - "Dog"

Subject: Edit Request for Wikipedia Entry

Subject: Request for Review and Clarification Regarding Draft Article

Subject: Concerns about Inaccurate Information

Subject: Behavioral issues and Wikihounding by User:Binksternet

AI chatbots sometimes generate messages with text broken into sections with titles. In some cases, such text may either appear with Markdown or as plain text when pasted. They may resemble or include inline-header vertical lists.

Importance of Thorough Research

Wikipedia's content guidelines emphasize the importance of verifiability and reliable sourcing, [...]

Risk of Violating Wikipedia’s Policy on Biographies of Living Persons

Wikipedia's Biographies of Living Persons (BLP) policy is particularly relevant in this context. [...]

The Dangers of Blindly Cutting Content

Blindly cutting content, especially when it pertains to significant aspects of a person’s public life, contradicts the core principles of Wikipedia. [...]

The Role of Due Diligence in Editing

Editors have a responsibility to engage in due diligence when making significant changes to articles, particularly those involving living persons. [...]

— From this August 2024 comment at Talk:Eric Dick (lawyer)

Response to Article Concerns

Source Independence and Coverage

The article has been updated to address concerns about source reliability. [...]

Swiss Media Coverage

The limited coverage in mainstream Swiss media likely reflects the niche nature of iolite Capital and its international investment focus. [...]

AI-Generated Content Suspicion

The perception that the article may be AI-generated seems to have been driven by previously listed dead or weak sources. [...]

— From this May 2025 comment at Wikipedia:Articles for deletion/Robert Leitz

Some LLM-generated comments contain level 2 or level 3 subheadings, or syntax intended to produce those subheadings, sometimes in Markdown, in which cases ## is used to denote each section.

Examples

In some cases, the syntax for subheadings appears in replies, where indentation prevents subheadings from forming.

=== Functional Wetland Ecology as Foundational Context ===

Regarding the "general" information on wetlands: I have included this as a foundational framework. [...]

=== Addressing Redundancy ===

Regarding the concern over "repetitive or cumulative" content: I appreciate this observation. [...]

— From this April 2026 comment at Talk:Great Black Swamp

== Request to Contest G11 Speedy Deletion of "Pharmaceutical Review" == I respectfully contest the speedy deletion tag (G11) placed on the article *Pharmaceutical Review*. [...]

---

== Appeal to Restore Submission: Consortium Publisher Canada - ICDTD Inc. ==

Dear KylieTastic,

Thank you for your thoughtful review of my submission for the article Consortium Publisher Canada - ICDTD Inc. [...]

== Submission Instructions for Restoration: ==

1. This message can be posted on your Talk Page under KylieTastic or the appropriate reviewer’s talk page.
2. Once revisions are made, submit the updated draft for Articles for Creation review again, ensuring all the required changes are made.

---

== Appeal Against Speedy Deletion Nomination of "Pakistan Pharmacist Federation" ==

Dear Helen,

Thank you for reviewing the "Pakistan Pharmacist Federation" article and for bringing the issue of its speedy deletion to my attention. [...]

=== 1. Notability of the Pakistan Pharmacist Federation ===

The Pakistan Pharmacist Federation (PFF) is a well-established national organization in Pakistan, [...]

=== 2. Compliance with Wikipedia’s Content Policies ===

I acknowledge the importance of adhering to Wikipedia’s policies, particularly the neutral point of view (NPOV) and verifiability guidelines. [...]

=== 3. Commitment to Citations from Reliable Sources ===

I understand that Wikipedia’s verifiability policy is crucial for maintaining the integrity of content. [...]

=== 4. Willingness to Collaborate with Experienced Editors ===

I understand the importance of following Wikipedia’s content policies and am open to collaborating with other experienced editors to improve the quality of the article. [...]

=== 5. Conclusion ===

Given the Pakistan Pharmacist Federation’s notability as a national organization with a significant role in public health and professional advocacy, [...]

---

== Appeal Against Speedy Deletion of "Advanced Multiple Incorporation Canada" ==

Dear Helen,

I appreciate your review and the notification regarding the speedy deletion nomination of the article "Advanced Multiple Incorporation Canada." [...]

=== 1. Notability of Advanced Multiple Incorporation Canada ===

Advanced Multiple Incorporation Canada (AMIC) is a registered Canadian company with a recognized presence within its industry, [...]

=== 2. Commitment to Adherence to Wikipedia’s Guidelines ===

I fully acknowledge that the initial draft may have lacked the required neutral tone and contained promotional language. [...]

=== 3. Third-Party Coverage and Reliable Sources ===

To meet Wikipedia’s standards for verifiability, I will update the article with citations from independent and reliable third-party sources, [...]

=== 4. Transparency and Improvements ===

I am committed to enhancing the article with relevant, accurate, and sourced information. [...]

=== 5. Commitment to Wikipedia’s Standards ===

It is important to emphasize that I respect Wikipedia’s editorial standards and guidelines, and I aim to contribute valuable and factual information to the platform. [...]

=== Conclusion ===

In light of the notability of AMIC, my willingness to rewrite the article with proper third-party sources, and my commitment to following Wikipedia’s content policies, [...]

— From these May 2026 comments at User talk:Tahanazir

When mentioning maintenance tags, AI chatbots often just write the names of such templates in curly brackets (e.g. {{Example}}), resulting in unintentional transclusions. These can be avoided by typing tl| between the opening pair of brackets and the template's name (e.g. {{tl|Example}}) so that each mention will instead appear like this: {{Example}}

Examples

From this September 2024 comment at Talk:2024 Bangladesh anti-Hindu violence

I would like to open a discussion regarding the recent edits and use of tags such as {{Unreliable sources}}, {{Disputed}}, and {{Cleanup rewrite}}. Additionally, the inclusion of claims about an “Indian disinformation campaign” requires careful consideration and sourcing in line with Wikipedia’s guidelines. [...] 3. Use of Maintenance Templates: Adding templates like {{Disputed}} and {{Unreliable sources}} should be done based on a consensus, not to preemptively frame content as problematic. I suggest we come to a consensus before continuing to apply these tags. Result I would like to open a discussion regarding the recent edits and use of tags such as Some of this article's listed sources may not be reliable. Please help improve this article by looking for better, more reliable sources. Unreliable citations may be challenged and removed. (Learn how and when to remove this message) , This article's factual accuracy is disputed. Relevant discussion may be found on the talk page. Please help to ensure that disputed statements are reliably sourced. (Learn how and when to remove this message) , and This article may need to be rewritten to comply with Wikipedia's quality standards. Relevant discussion may be found on the talk page. You can help. The talk page may contain suggestions. . Additionally, the inclusion of claims about an “Indian disinformation campaign” requires careful consideration and sourcing in line with Wikipedia’s guidelines. [...] 3. Use of Maintenance Templates: Adding templates like This article's factual accuracy is disputed. Relevant discussion may be found on the talk page. Please help to ensure that disputed statements are reliably sourced. (Learn how and when to remove this message) and Some of this article's listed sources may not be reliable. Please help improve this article by looking for better, more reliable sources. Unreliable citations may be challenged and removed. (Learn how and when to remove this message) should be done based on a consensus, not to preemptively frame content as problematic. I suggest we come to a consensus before continuing to apply these tags.

From this July 2025 comment at Wikipedia:Articles for deletion/AI book generation (2)

Regarding [[WP:TNT]], you've misrepresented it. TNT specifically states it's for "unsalvageable articles" and explicitly notes "If the problems are limited to a section or two, {{section rewrite}} tags are more appropriate." An article with 80+ references addressing a notable topic hardly qualifies as "unsalvageable." Result Regarding WP:TNT, you've misrepresented it. TNT specifically states it's for "unsalvageable articles" and explicitly notes "If the problems are limited to a section or two, This article may need to be rewritten to comply with Wikipedia's quality standards. You can help. The talk page may contain suggestions. tags are more appropriate." An article with 80+ references addressing a notable topic hardly qualifies as "unsalvageable."

From this June 2026 comment at Talk:Culture jamming

I removed the article-wide {{Technical}} and {{Essay-like}} maintenance tags after a conservative copyedit of the lead, history, tactics, and criticism sections. The edits were intended to improve accessibility and encyclopedic tone while retaining the article's structure, examples, and existing references. The article still contains some theoretical terminology, such as détournement, metameme, and late capitalism, but these terms are now presented in a more explanatory and attributed way rather than as an argumentative essay. The criticism section was also expanded to address co-optation, commodification, symbolic dissent, and related approaches. Result I removed the article-wide This article may be too technical for most readers to understand. Please help improve it to make it understandable to non-experts, without removing the technical details. (Learn how and when to remove this message) and This article is written like a personal reflection, personal essay, or argumentative essay that states a Wikipedia editor's personal feelings or presents an original argument about a topic. Please help improve it by rewriting it in an encyclopedic style. (Learn how and when to remove this message) maintenance tags after a conservative copyedit of the lead, history, tactics, and criticism sections. The edits were intended to improve accessibility and encyclopedic tone while retaining the article's structure, examples, and existing references. The article still contains some theoretical terminology, such as détournement, metameme, and late capitalism, but these terms are now presented in a more explanatory and attributed way rather than as an argumentative essay. The criticism section was also expanded to address co-optation, commodification, symbolic dissent, and related approaches.

Links to searches

• Special:WhatLinksHere/Template:AI-generated (talk pages only, links and redirects hidden)

When a user is blocked, they might ask an AI chatbot to write an unblock request for them. Many AI-generated unblock requests tend to go something like this:

Dear Wikipedia Administrators,
I respectfully request that the block on my account be lifted. I acknowledge that my recent behavior has been in violation of Wikipedia's standards. My intention was to edit constructively, and I sincerely apologize for any disruption I may have caused. I have taken the time carefully read and review all of Wikipedia's policies and guidelines, and I now understand the importance of adhering to them when editing. Moving forward, I am committed to ensuring that all of my future edits are constructive. If there are any specific concerns that need further clarification, feel free to ask me about them, and I will be happy to address them. I look forward to collaborating in a constructive manner. Thank you for your time and consideration.

In these requests, blocked users often try to assure administrators of their good faith and (future) adherence to policies, express their willingness to communicate, and request feedback to help them improve their editing.

Large language models have been used to write reports at Wikipedia:Administrators' noticeboard/Incidents, as evidenced by the presence of indicators that have appeared on LLM-generated text elsewhere on Wikipedia. These reports often present each issue as entries on a list or even separate sections divided by subheadings, whose titles are written in title case. Some of them also use boldface, or attempt to do so via Markdown.

Examples

1. **Account Age:** [...]

2. **Prior SPI Context:** [...]

3. **Exact Lead Quote Re-insertion:** [...]

4. **Active Cross-Page Disruption:** [...]

Sealioning is a disruptive practice where someone feigns ignorance and relentlessly demands users to provide answers or evidence to back up the statements they make. When targets express frustration over the sealioner's behavor, the sealioner criticizes the targets for responding negatively to them, even though they were just asking questions politely, or so they present themselves as polite.

Several AI users have resorted to sealioning when confronted about their AI use. They write long comments that sometimes present arguments as entries in bulleted lists or separated into multiple sections. LLM users often ask for examples of problematic text or stronger evidence than "stylistic indicators", and accuse their confronters of acting on unsubstantiated speculation and being uncivil when they don't get the answers they want.

Wikilawyering is a disruptive practice where someone selectively cites or interprets policies, guidelines, or perceived precedent as justification for their conduct, even if their interpretations go against the purpose of the policies or guidelines they mention. In many cases, users have emphasized their content's compliance (or overlooked their non-compliance) with certain policies and guidelines. This is especially the case for users of AI chatbots, which often generate text that affirms what the user may want others to believe, even if the points they present don't actually hold up against Wikipedia's policies and guidelines.

When an article is tagged as possibly containing AI-generated content, a user might try to defend it by asking accusers to point to specific passages causing concern or reassuring them that the content they've contributed is "neutral", "verified" by citations to reliable sources, and comprises none of the items described in Template:AI-generated, regardless of the amount of actual effort put in to ensure that such claims are true.

Common AI wikilawyering tropes include:

• Affirming that one's content adheres to Wikipedia's policies and guidelines. A few core content policies—such as Neutral point of view, Verifiability, Reliable sources, and No original research—are often itemized by LLM users.
• Emphasizing that Wikipedia's LLM guideline allows AI to be used for basic copyediting or translation.
• Outright denying that Wikipedia prohibits using AI to write or rewrite content at all.
• Claiming that accusations of AI use must be backed by hard evidence asides from stylistic indicators. In some cases, WP:AIDETECTIVE has been cited.

On a couple of occasions, AI users have claimed that Wikipedia's guidance on LLM use permits the use of LLMs to "spot structural problems", "generate ideas" or for help "improv[ing] a paragraph".^[c] Although these specific words do not appear at WP:LLM (which they erroneously cite when mentioning these exceptions), they do appear in this paragraph at Wikipedia:Large language models § Writing articles, which may be what they mean to refer to:

The ban does not cover indirect use of LLMs, which means that they can be used adjacent to editing. For example, they can help editors spot structural problems in long articles (such as: the same information being repeated in different sections, the infobox not agreeing with the prose, etc.), and to generate ideas for new or existing articles. If using an LLM as a writing advisor, i.e. asking for an outline, how to improve a paragraph, a critique of existing content, etc., editors should remain aware that the information it gives is unreliable. Due diligence and common sense are required when choosing whether to incorporate any suggestions. The editor should become familiar with the sourcing landscape for the topic in question and then carefully evaluate the suggestions for their neutrality and verifiability.

False accusations of AI use can drive away new editors and foster an atmosphere of suspicion. Before claiming AI was used, consider whether the Dunning–Kruger effect or confirmation bias may be clouding your judgement. Detecting LLM texts on the basis of style alone is not as easy as it seems, see WP:AIDETECTIVE. Here are some somewhat commonly-used indicators that are ineffective in LLM detection—and may even indicate the opposite.

• Letter-like writing (in isolation) – Although many talk page messages written with salutations, valedictions, subject lines, and other formalities after 2023 tend to appear AI-generated, letters and emails have conventionally been written in such ways long before modern LLMs existed. Human editors (particularly newer editors) may format their talk page comments similarly for various reasons, such as being more accustomed to formal communication, posting as part of a school assignment that requires this tone, or simply mistaking the talk page for email. Other tells, such as vertical lists, placeholders, or abrupt cutoffs, are stronger.

• Requests for source assessment (in isolation) – Although a few LLM users who have their drafts declined may go to the AfC help desk to ask for an editor (in some cases, an "experienced" or "uninvolved" one) to advise them as to which sources (out of the ones they're trying to use) help establish the notability of a given topic,^[d] such inquiries have also been made by users who don't use LLMs. In some cases, a user might be trying to figure out how certain cells on a particular source assessment table should be filled.

• Wikipedia:Help, I've been accused of AI!
• Wikipedia:Yes, you have to follow NOLLM
• Wikipedia:Yes, that does violate the AI guidelines
• User:LWG/These are the LLM users in your neighborhood