Wikipedia says AI answers are starting to take a bite. There are reasons to be worried. (www.businessinsider.com)

Wikipedia says global pageviews are down about 8% year‑over‑year in recent months, and it blames a big part of that dip on AI chatbots and search engines that extract and repackage Wikipedia content without sending users to the site. The result: people get quick answers (e.g., contradictory snippets about the White House East Wing) without visiting Wikipedia, which threatens the nonprofit’s ecosystem—fewer casual visits can mean fewer donors and fewer potential new editors—even as Wikimedia stresses that wider distribution of its “free knowledge” is mission-aligned. Other factors—like younger users getting info from video—also play a role, and the site still draws over 10 billion monthly views and raised $170.5M last year. Technically, the story matters because Wikipedia is a core, high‑quality dataset for training and grounding large language models; studies show model outputs drop in quality when Wikipedia is excluded. That gives Wikimedia leverage to monetize access via Wikimedia Enterprise’s API, though many models have already ingested huge amounts of Wikipedia content for free. The tension is now between raising revenue/attribution and preserving the deep browsing experience (the “Wikipedia rabbit hole”) that drives editing and donations. Expect more licensing deals and political scrutiny as platforms, publishers and Wikimedia negotiate how AI uses human‑curated knowledge.