The impact of AI and fake news on Wikipedia (diff.wikimedia.org)

At WikiCon in Potsdam this October—the largest gathering of German‑language Wikipedia volunteers (nearly 400 attendees)—participants tackled how generative AI is reshaping access to and the integrity of the free encyclopedia. Presenters warned that large language models increasingly rely on Wikipedia as training data even as younger users shift to AI-driven search: visits from 18–24 year‑olds fell from 30% in 2019 to 8% today. That loop—AI consuming Wikipedia while AI tools divert readers—creates both opportunity (Wikipedia remains a crucial free knowledge source) and risk (less direct engagement, shifting expectations for content format and discoverability). Volunteers are already confronting a spike in AI‑generated submissions that often contain fabricated facts, bogus citations, and non‑encyclopedic language. Harald Krichel outlined practical “hard” and “soft” signals to spot such articles: non‑existent publications or ISBNs, leftover chat phrases, fake or invalid infobox fields, dead or fabricated links, wikitext/markup remnants, odd structural patterns (excessive lists, emojis), flowery/promotional prose, and sudden prolific newcomers. Detected contributions are either revised to meet verifiability and neutral‑tone standards or deleted. The community and Wikimedia Deutschland emphasized the wider stakes: as disinformation, deepfakes, and AI‑cloned news sites proliferate, rigorous source‑checking and community moderation are essential to preserve Wikipedia’s role as a verifiable counterweight to algorithmically amplified misinformation.