Wikipedia: WikiProject AI Cleanup

Wikipedia’s new AI Cleanup Crew sparks a fight over “slop,” rules, and a pre-ChatGPT time machine

TLDR: Wikipedia launched an “AI Cleanup” project to find and fix unsourced, error-prone AI text and citations. Commenters are split between cheering clearer rules, demanding a pre-ChatGPT “time machine,” and warning that “it sounds like AI” isn’t proof—raising big trust questions for the web’s biggest reference site.

Wikipedia just rolled out a volunteer “AI Cleanup” squad to hunt down sloppy, unsourced chatbot prose and fake citations, and the comments lit up. Supporters cheered the promise of receipts over robot vibes, pointing to the detailed Signs of AI writing guide—one fan said it goes way beyond the old “em-dashes” meme. Others wanted the energy pointed both ways: fix AI junk, yes, but also rewrite the legendary human-written gobbledygook that’s been confusing readers for years. The spiciest request? A pre-ChatGPT “time machine,” like the Coppermind time-travel, so readers can browse a slop-free snapshot when they’re spoiler- or bot-averse.

Not a ban, the project aims to verify or delete AI-made text and images—especially pages with fake sources or whole hoaxes (one AI-written article even invented Russian and Hungarian citations; another beetle entry cited off-topic French and German pieces; an entire fake place got nuked). Oh, and editors can even speedy-delete obvious bot pages. Skeptics pushed back, warning that “it sounds like AI” isn’t proof, and that witch-hunting style cues risks tossing good edits. One commenter also tied it to Wikipedia’s business reality: partnerships depend on human-made knowledge. Verdict from the crowd: protect sources, publish the telltale signs, but give us tools, rules—and maybe that time portal

Key Points

  • Wikipedia launched WikiProject AI Cleanup to address unsourced and error-prone AI-generated content.
  • The project targets AI-written text and AI-generated images, emphasizing verification rather than banning AI.
  • Editors are advised to tag, remove unsourced material, warn users, and nominate wholly LLM-generated pages for speedy deletion under WP:G15.
  • Detection guidance includes a “Signs of AI writing” resource and noting that pre–Nov 30, 2022 content is unlikely to be AI-generated.
  • Examples show fabricated or irrelevant citations and a hoax article (Amberlihisar) created in 2023 and later deleted.

Hottest takes

"It goes into a lot more detail than the typical \"em-dashes\" heuristic" — Antibabelic
"I would like a similar pre-LLM Wikipedia snapshot" — progbits
"'It sounds like AI' is not a good metric whatsoever" — weli
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.