Wikipedia's Cleanup of AI-Generated Content

Wikipedia’s volunteer WikiProject AI Cleanup began in 2023 after editors noticed articles displaying recognizable chatbot habits, including repeated clichés, unnecessary conclusions, and “why is this important?” sections. As AI writing has become less obvious, editors now look for accumulated stylistic patterns - such as repeated use of three-part lists and contrastive phrasing - as well as suspicious editing behavior, like someone producing several long articles in minutes. Misleading citations are a major concern: AI may cite a real source while misrepresenting what it says. English Wikipedia now prohibits editors from using large language models to generate articles, but the scale of the problem remains uncertain. A detection company flagged about 20 percent of new articles and drafts in late 2025, falling to 15-16 percent in early 2026; that change could reflect the policy or improving AI that is harder to detect.

Editors are also testing AI as a cleanup aid. One tool checks whether sources support page claims, while Wikimedia’s Paste Check aims to identify text copied from some chatbots. Lebleu says the ban was introduced because editors lacked tools to handle a flood of machine-generated material, but the policy has come to express Wikipedia’s identity as a human-written encyclopedia. Whether that identity should endure if AI can produce reliable articles remains unsettled; one possibility is preserving human-written Wikipedia while offering a separate AI-assisted option. Lebleu also sees a risk that AI search could cite Wikipedia without bringing readers to the site or inviting them to edit. Wikimedia is exploring ways to show citation context and enable corrections, suggesting a future in which AI both challenges Wikipedia and helps people use and maintain it.