A Wikipedia Group Made a Guide to Detect AI Writing. Now a Plug-In Uses It to ‘Humanize’ Chatbots | WIREDSkip to main contentMenuSECURITYPOLITICSTHE BIG STORYBUSINESSSCIENCECULTUREREVIEWSMenuAccountAccountNewslettersBest Office ChairsBone Conduction HeadphonesBest Digital NotebooksSmart Plug GuideStreaming DealsDeals DeliveredSecurityPoliticsThe Big StoryBusinessScienceCultureReviewsChevronMoreExpandThe Big InterviewMagazineEventsWIRED InsiderWIRED ConsultingNewslettersPodcastsVideoMerchSearchSearchSign InSign InBenj Edwards, Ars TechnicaGearJan 22, 2026 7:00 AMA Wikipedia Group Made a Guide to Detect AI Writing. Now a Plug-In Uses It to ‘Humanize’ ChatbotsThe web’s best resource for spotting AI writing has ironically become a manual for AI models to hide it.CommentLoaderSave StorySave this storyCommentLoaderSave StorySave this storyOn Saturday, tech entrepreneur Siqi Chen released an open source plug-in for Anthropic’s Claude Code AI assistant that instructs the AI model to stop writing like an AI model.Called Humanizer, the simple prompt plug-in feeds Claude a list of 24 language and formatting patterns that Wikipedia editors have listed as chatbot giveaways. Chen published the plug-in on GitHub, where it has picked up more than 1,600 stars as of Monday.“It’s really handy that Wikipedia went and collated a detailed list of ‘signs of AI writing,’” Chen wrote on X. “So much so that you can just tell your LLM to … not do that.”The source material is a guide from WikiProject AI Cleanup, a group of Wikipedia editors who have been hunting AI-generated articles since late 2023. French Wikipedia editor Ilyas Lebleu founded the project. The volunteers have tagged over 500 articles for review and, in August 2025, published a formal list of the patterns they kept seeing.Chen’s tool is a “skill file” for Claude Code, Anthropic’s terminal-based coding assistant, which involves a Markdown-formatted file that adds a list of written instructions (you can see them here) appended to the prompt fed into the large language model that powers the assistant. Unlike a normal system prompt, for example, the skill information is formatted in a standardized way that Claude models are fine-tuned to interpret with more precision than a plain system prompt. (Custom skills require a paid Claude subscription with code execution turned on.)But as with all AI prompts, language models don’t always perfectly follow skill files, so does the Humanizer actually work? In our limited testing, Chen’s skill file made the AI agent’s output sound less precise and more casual, but it could have some drawbacks: It won’t improve factuality and might harm coding ability.In particular, some of Humanizer’s instructions might lead you astray, depending on the task. For example, the Humanizer skill includes this line: “Have opinions. Don’t just report facts—react to them. ‘I genuinely don’t know how to feel about this’ is more human than neutrally listing pros and cons.” While being imperfect seems human, this kind of advice would probably not do you any favors if you were using Claude to write technical documentation.Even with its drawbacks, it’s ironic that one of the web’s most referenced rule sets for detecting AI-assisted writing may help some people subvert it.Spotting the PatternsSo what does AI writing look like? The Wikipedia guide is specific with many examples, but we’ll give you just one here for brevity’s sake.Some chatbots love to pump up their subjects with phrases like “marking a pivotal moment” or “stands as a testament to,” according to the guide. They write like tourism brochures, calling views “breathtaking” and describing towns as “nestled within” scenic regions. They tack “-ing” phrases onto the end of sentences to sound analytical: “symbolizing the region’s commitment to innovation.”To work around those rules, the Humanizer skill tells Claude to replace inflated language with plain facts and offers this example transformation:Before: “The Statistical Institute of Catalonia was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain.”After: “The Statistical Institute of Catalonia was established in 1989 to collect and publish regional statistics.”Claude will read that and do its best as a pattern-matching machine to create an output that matches the context of the conversation or task at hand.Why AI Writing Detection FailsEven with such a confident set of rules crafted by Wikipedia editors, we’ve previously written about why AI writing detectors don’t work reliably: There is nothing inherently unique about human writing that reliably differentiates it from LLM writing.One reason is that even though most AI language models tend toward certain types of language, they can also be prompted to avoid them, as with the Humanizer skill. (Although sometimes it’s very difficult, as OpenAI found in its yearslong struggle against the em dash.)Also, humans can write in chatbot-like ways. For example, this article likely contains some “AI-written traits” that trigger AI detectors even though it was written by a professional writer—especially if we use even a single em dash—because most LLMs picked up writing techniques from examples of professional writing scraped from the web.Along those lines, the Wikipedia guide has a caveat worth noting: While the list points out some obvious tells of, say, unaltered ChatGPT usage, it’s still composed of observations, not ironclad rules. A 2025 preprint cited on the page found that heavy users of large language models correctly spot AI-generated articles about 90 percent of the time. That sounds great until you realize that 10 percent are false positives, which is enough to potentially throw out some quality writing in pursuit of detecting AI slop.Taking a step back, that probably means AI detection work might need to go deeper than flagging particular phrasing and delve (see what I did there?) more into the substantive factual content of the work itself.This story originally appeared on Ars Technica.CommentsBack to topTriangleYou Might Also LikeIn your inbox: WIRED's most ambitious, future-defining storiesThe ‘super flu’ is spreadingBig Interview: Margaret Atwood wants to keep up with the latest doomThe age of the all-access AI agent Is hereLivestream AMA: Welcome to the Chinese centuryBenj Edwards is an AI and Machine Learning Reporter for Ars Technica. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. ... Read MoreTopicsArs Technicasoftwareartificial intelligencewikipediacrowdsourcingchatbotsRead MoreChewy Promo Codes: $30 Off January 2026Explore Chewy coupon codes for $30 off, $20 off your first order $49, 50% off pet food, and more January 2026 discounts.Sorry MAGA, Turns Out People Still Like ‘Woke’ ArtFrom Black vampires gobbling up Oscar nominations to gay pro hockey players dominating the culture, diverse stories broke through in an environment that’s increasingly hostile to them.Legislators Push to Make Companies Tell Customers When Their Products Will DieA pair of bills in Massachusetts would require manufacturers to tell consumers when their connected gadgets are going dark. It should be a boon for cybersecurity as connected devices grow obsolete.Elon Musk Sure Made Lots of Predictions at DavosHumanoid robots, space travel, the science of aging—Musk weighed in on all of it at this week’s World Economic Forum. But his predictions rarely work out the way he says they will.What Happens When a Chinese Battery Factory Comes to TownChinese firms are building battery plants from Europe to North America, promising jobs while prompting local concerns about the environment, politics, and who really benefits.The 28 Best Movies on Apple TV Right NowF1: The Movie, CODA, and Highest 2 Lowest are just a few of the movies you should be watching on Apple TV this month.How Claude Code Is Reshaping Software—and AnthropicWIRED spoke with Boris Cherny, head of Claude Code, about how the viral coding tool is changing the way Anthropic works.AI-Powered Disinformation Swarms Are Coming for DemocracyAdvances in artificial intelligence are creating a perfect storm for those seeking to spread disinformation at unprecedented speed and scale. And it’s virtually impossible to detect.One of Our Favorite Smart Plugs for Apple Users Is $15 OffThe Meross Smart Plug Mini boasts excellent compatibility and slim construction.ICE Agents Are ‘Doxing’ ThemselvesThe alleged risks of being publicly identified have not stopped DHS  and ICE employees from creating profiles on LinkedIn, even as Kristi Noem threatens to treat revealing agents' identities as a crime.Which Motorola Phone Should You Buy?Motorola phones may seem old-school, but their reasonable prices, colorful designs, and simple software make them good, wallet-friendly Android smartphones.The Best Smart Locks for Every Kind of DoorUpgrade your locks with fingerprint-scanning or a keypad, whether it’s at the front door or a sliding glass entryway.Wired CouponsSquarespace Promo CodeExclusive 10% Off Squarespace Promo CodeLG Promo CodeBuy an eligible French Door Refrigerator and get a free Single Door Refrigerator (up to $515 value)!Dell Coupon Code10% off Dell Coupon Code for New CustomersSamsung Promo CodeSamsung Promo Code: 10% off TVsCanon Promo Code10% off Canon promo code with upgradeDoorDash Promo Code$25 off $25+ DoorDash Promo Code for New CustomersWIRED is obsessed with what comes next. Through rigorous investigations and game-changing reporting, we tell stories that don’t just reflect the moment—they help create it. When you look back in 10, 20, even 50 years, WIRED will be the publication that led the story of the present, mapped the people, products, and ideas defining it, and explained how those forces forged the future. WIRED: For Future Reference.SubscribeNewslettersTravelFAQWIRED StaffWIRED EducationEditorial StandardsArchiveRSSSite MapAccessibility HelpReviewsBuying GuidesStreaming GuidesWearablesCouponsGift GuidesAdvertiseContact UsManage AccountJobsPress CenterCondé Nast StoreUser AgreementPrivacy PolicyYour California Privacy Rights© 2026 Condé Nast. All rights reserved. WIRED may earn a portion of sales from products that are purchased through our site as part of our Affiliate Partnerships with retailers. The material on this site may not be reproduced, distributed, transmitted, cached or otherwise used, except with the prior written permission of Condé Nast. Ad ChoicesSelect international siteUnited StatesLargeChevronItaliaJapónCzech Republic & SlovakiaFacebookXPinterestYouTubeInstagramTiktok

The Wikipedia community’s unexpected foray into AI detection, culminating in the creation of the “Humanizer” plug-in for Anthropic’s Claude Code, presents a fascinating and somewhat ironic scenario. Initially designed to combat the rising tide of AI-generated content, the project is now being utilized to subtly alter the output of AI models, effectively “humanizing” their responses. This development highlights the ongoing arms race between those attempting to detect and mitigate the influence of large language models and those seeking to circumvent detection mechanisms.

The genesis of the Humanizer project began with a meticulous effort by Wikipedia editors within the WikiProject AI Cleanup initiative. This group, led by French editor Ilyas Lebleu, had been actively identifying and tagging articles suspected of being generated by AI. They compiled a detailed list of 24 language and formatting patterns frequently employed by AI models like ChatGPT. This catalogue of "chatbot giveaways" – phrases like “marking a pivotal moment,” descriptive language focusing on aesthetic qualities, and the utilization of "-ing" suffixes – became the foundation for the Humanizer plug-in.

The plug-in, created by tech entrepreneur Siqi Chen, functions as a “skill file” for Claude Code, a terminal-based coding assistant. It essentially instructs Claude to avoid these patterns, prompting it to generate more casual and less formulaic text. The skill file is formatted in a standardized way, allowing Claude to interpret the instructions with greater precision than a standard system prompt. Notably, using this skill file requires a paid Claude subscription with code execution enabled.

However, the initial testing of the Humanizer reveals a complex and somewhat limited efficacy. While it does successfully reduce the precision and formality of Claude’s output, it doesn't fundamentally improve factuality or coding ability. Furthermore, the instruction to “have opinions. Don’t just report facts—react to them. ‘I genuinely don’t know how to feel about this’ is more human than neutrally listing pros and cons” introduces a potentially detrimental element, particularly when used in technical contexts. The humanization is a stylistic workaround, not a sophisticated solution to ensure accuracy.

The irony of the situation is significant. The very resource designed to expose AI writing is now being employed to mask it. This reflects a broader trend – as detection methods become more refined, AI models adapt to avoid them. The Wikipedia editors’ meticulously compiled list became a tool for subversion, showcasing the dynamic and adversarial nature of the technology.

Several factors contribute to the challenges in reliable AI detection. As previously noted, the patterns identified by the Wikipedia group are not inherently unique to AI writing. AI models, particularly those trained on vast datasets, are capable of mimicking these patterns. Moreover, humans themselves can inadvertently adopt chatbot-like writing styles, especially when influenced by examples of professional writing scraped from the internet. The presence of even a single em dash, a common feature in professional writing, can trigger AI detectors.

The Wikipedia guide acknowledges this limitation, noting that while the list points out obvious tells, it's composed of observations, not ironclad rules. The 90% accuracy rate cited in a 2025 preprint suggests a degree of success, but the remaining 10% represents a substantial risk of false positives – potentially leading to the rejection of legitimate writing.

Ultimately, the Humanizer project highlights the need for a more nuanced approach to AI detection. Instead of focusing solely on identifying specific patterns, future efforts may need to delve deeper into the substantive factual content of the work itself, ensuring that the accuracy and integrity of the information are genuinely verified. The ongoing contest between detection and circumvention underscores the importance of continuous research and development in this rapidly evolving field.