AI Is Pulling Readers From Wikipedia, Which Its Co-founder Says AI Still Relies On

As Wikipedia marked 25 years since its founding, interviewer Johnny Diamond asked co-founder Jimmy Wales whether artificial intelligence could disrupt the volunteer written encyclopedia. Wales said traffic had slipped for short answers, as people asked chatbots and Google for quick facts. Then he gave his warning to AI companies.

"The only reason your AI is pretty good is because it's read all of Wikipedia," Wales says.

He describes Wikipedia's contributors as "a community of nerds who write an encyclopedia as our hobby and we just enjoy doing it." That hobby, he says, supplies what he called the "human layer of judgment and taste and thoughtfulness" that chatbots depend on. Without it, he says, AI systems would be left to train on poor material and would end up, in his words, "pretty angry and stupid."

The pattern is already visible, Wales says. People who once typed a quick question into Google now take a short answer from AI, which he said he had read was hurting Google's ad business. But readers still turn to Wikipedia when they want to truly understand a subject. When Dolly Parton was in the news, he says, millions came to read her full life story and background in a way no brief chatbot reply can match.

Wales does not dismiss the threat. He says it is not hard to imagine AI improving, hallucinating less, and becoming a competitor in some sense. His concern is what happens if firms train future systems largely on the output of current ones. He points to industry talk of what researchers call model collapse, where systems trained only on AI output spiral into nonsense.

That is why, he says, human checking still matters. A reporter who witnessed an event, or a volunteer editor who studied the coverage and decided what was reliable, can help ground AI in what he calls ground truth.