Wikipedia's Trusted Data Feeds Chatbot Misinformation Pipeline

Wikipedia became the internet's default source of factual authority. Now, the same pattern is repeating with artificial intelligence chatbots—with significantly higher stakes. Major AI systems including ChatGPT, Claude, Gemini, and Grok are trained heavily on Wikipedia's human-curated content, which has become a foundational pillar of modern language models. Yet these same systems struggle with factual accuracy: a Bloomberg study found that answers about elections and geopolitics failed on accuracy, bias, or proper sourcing 90% of the time. The transitive trust—from Wikipedia to chatbots to users—creates a dangerous cycle where fluent, confident language from AI systems masks underlying factual errors, researchers warn.