Wikipedia is written and policed by volunteers, and after ChatGPT made fluent text generation free and instant, those volunteers started finding articles that had been quietly written or padded by chatbots. In December 2023, a group of editors organized WikiProject AI Cleanup, a collaboration “aimed at dealing with AI-generated content on Wikipedia.” Its goal was not to ban AI outright but to “identify and address the misuse of AI in articles” - removing unsourced claims, flagging fabricated references, spotting AI-generated images, and educating editors about the technology’s limits.
The problem the project targets is specific: large language models produce text that reads authoritatively but often invents facts, citations, and sources. As the project notes, this is hard to police because “generated text is often indistinguishable from human text.” To help, editors compiled catalogs of telltale signs - leftover chatbot phrasing such as “as of my last knowledge update,” hedging boilerplate, and fabricated citations that point to nonexistent papers.
The cleanup effort sits alongside Wikipedia’s broader guidance on large language models, which by the mid-2020s discouraged using LLMs to generate or rewrite article content outright, allowing only narrow exceptions such as translation or basic copyediting where a human verifies the result. The community’s stance reflected hard experience: unverified machine text was a net cost to an encyclopedia whose entire value rests on sourcing and accuracy.
Why business readers should care: WikiProject AI Cleanup is a window into what happens when generative AI meets a system that depends on verifiable sourcing. The volunteer response - detection heuristics, human verification, and a policy that treats unsourced AI text as a liability rather than a labor saving - mirrors the controls serious organizations end up adopting.