Wikipedia bans AI-generated articles after editor revolt

English Wikipedia now prohibits editors from using generative AI to write or rewrite entire articles. The policy, added to Wikipedia's guidelines on March 21, blocks any content "entirely generated" by large language models. Two narrow exceptions remain: editors can use AI for translation assistance and for generating short snippets during routine edits, but must personally verify all facts before publishing .

The ban follows months of internal debate where editors flagged repeated policy violations from AI content. According to the Wikimedia Foundation, AI-generated text "often violates several of Wikipedia's core content policies" around verifiability, neutral point of view, and reliable sourcing. The final straw appears to have been an influx of poorly-written articles with hallucinated citations that required extensive cleanup by human volunteers .

The policy change passed with overwhelming support. Over 40 editors participated in the formal discussion, with the measure receiving broad consensus for immediate implementation. Volunteer moderators have since launched "WikiProject AI Cleanup" to systematically identify and remove existing AI slop. New speedy deletion rules (code G15) let admins instantly nuke articles showing clear LLM fingerprints like fake citations or robotic phrasing .

Wikipedia isn't conducting a mass purge, but new AI detection workflows are now in place. Articles flagged as potentially AI-generated get reviewed by human editors within 48 hours. If confirmed as AI-written, they face deletion regardless of factual accuracy. The community's stance: even correct AI content fails Wikipedia's standards because it lacks traceable human authorship and editorial judgment .

The policy currently covers English Wikipedia only, but momentum is building for wider adoption. German and French Wikipedia communities are already discussing similar measures. Spanish and Japanese editions appear more hesitant, citing translation needs for underrepresented topics. The Wikimedia Foundation has signaled it won't impose blanket rules, leaving each language edition to decide independently .

The ban represents a major reputational blow for generative AI vendors. Wikipedia's decision undercuts key marketing claims about AI assisting knowledge work. OpenAI, Google, and Anthropic have all promoted LLMs as research assistants, but Wikipedia's rejection suggests fundamental quality issues remain. Smaller AI writing startups that targeted academic markets face immediate customer loss as institutions follow Wikipedia's lead .

Enforcement relies on human pattern recognition rather than automated detection. Editors are trained to spot AI giveaways: unusual citation patterns, generic prose, or factual inconsistencies. The community explicitly rejected AI detection tools as unreliable, noting that false positives could harm legitimate editors. Repeat violators risk losing editing privileges, though first-time offenders typically get warnings .

This sets a precedent for other collaborative knowledge bases. Stack Overflow is reviewing similar restrictions. Academic publishers like Nature and Science are watching closely, with some considering mandatory AI disclosure requirements. The move reinforces a growing consensus: human oversight remains essential for maintaining trust in public information repositories .