Wikipedia has updated its editorial guidelines to explicitly prohibit editors from using artificial intelligence tools to write or rewrite entry content. The change was added to Wikipedia pages over the weekend on the grounds that AI-generated articles often violate many of Wikipedia's core content policies.
The new rules currently apply to the English version of Wikipedia, but do not completely negate all uses of AI. Editors are still allowed to use large language models in limited scenarios, such as to make basic grammar and wording modification suggestions for existing texts, but the premise is that the AI cannot introduce any new content on its own. In addition, editors can use AI to translate Wikipedia entries in other languages into English, but they must abide by the website's existing rules on "LLM-assisted translation" and have sufficient original language proficiency to verify the accuracy of the translation.
In response to possible "style witch hunts" from the outside world, the new guidelines specifically remind that some users' writing styles themselves may be similar to large language models. Therefore, if you want to restrict an editor's editing rights based on suspicion of AI abuse, you should not rely solely on stylistic or linguistic features, but should focus on whether the text complies with Wikipedia's core content policies and make a comprehensive judgment based on its recent editing records.
Prior to this, the Wikipedia community had been "dealing" with AI-generated content for several months and introduced a series of countermeasures. The community has adopted a new policy that allows for "quick deletion" of low-quality AI-generated entries to reduce the erosion of "AI irrigation content" on the quality of the entire platform. Editors have also formed a project called "WikiProject AI Cleanup" to clean up content written by AI and assist other editors in identifying such text.
The blanket ban on writing articles using large language models is based on a proposal initiated by a community member called Chaotic Enby. The proposal sparked lengthy discussions within the community and ultimately received “overwhelming support”. In summarizing the reasons for adoption, the community stated that the new policy targets the most problematic use cases of large language models, but still leaves room for some AI assistance functions that are considered "reasonable uses."