Wikipedia has signed deals with Amazon, Meta, Microsoft, Perplexity, and Mistral AI. The nonprofit that famously rejected advertising is now a paid data supplier to the AI industry. Google signed in 2022; the rest just caught up.
Wikipedia’s human traffic dropped 8% year-over-year, according to data the Wikimedia Foundation published in October 2025. Research from Profound analyzing 680 million AI citations found that Wikipedia accounts for 47.9% of ChatGPT’s top-10 most-cited sources. As you know, generative AI models and reasoning engines answer questions by summarizing Wikipedia content (often without citing, or linking to, the source).
Wikipedia hosts 65 million articles in 300 languages with 15 billion monthly page views. It’s the only nonprofit in the top 10 most-visited websites. Selena Deckelmann, Wikimedia’s CPO/CTO, said, “Wikipedia shows that knowledge is human, and knowledge needs humans. Especially now, in the age of AI, we need the human-powered knowledge of Wikipedia more than ever.”
I’m fascinated by Wikipedia’s role in AI training. In my experience with the site, when a Wikipedia entry gets it right, it’s usually very right. But, when an entry is wrong, it’s worse than an AI hallucination in that someone intentionally entered the wrong info and the WikiPolice (a volunteer army of human editors) either missed it or let it go. This is, of course, armchair research and one person’s anecdotal experience. I’m just wondering how AI models will vet the information.
As for the deal, costs less than litigation and structured data access beats scraping. Wikimedia Enterprise gives AI companies what they need at the volume and speed they require. Call it infrastructure maintenance. This is also a good time to reflect on the fact that your favorite foundation models depend on data sources you don’t control.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.