Relatively few Americans are getting news from AI chatbots like ChatGPT (www.pewresearch.org)

A Pew Research Center survey finds that while a growing number of Americans have tried AI chatbots like ChatGPT and Gemini, they remain a minor news source: only 2% say they get news often and 7% sometimes from chatbots, 16% rarely, and 75% never. Fewer than 1% prefer chatbots over other news outlets. Adults under 50 are modestly more likely to use chatbots for news (12% vs. 6% for older adults), but overall adoption as a news channel is limited. Users report mixed experiences: one-third say it’s generally difficult to judge truthfulness, 24% find it easy, and 42% are unsure. About half of chatbot-news consumers encounter information they think is inaccurate at least sometimes (16% say often/extremely often), with younger users (59% of 18–29) more likely to report encountering inaccuracies than older groups. For the AI/ML community, these findings underscore that chatbots are not yet trusted news intermediaries and highlight pressing technical and product priorities: reducing hallucinations, improving provenance and source attribution, better uncertainty calibration and explainability, and UX that helps users verify claims. The age-skew in perceived inaccuracy suggests models and interfaces should be tailored to younger heavy users and that evaluation metrics should prioritize factuality in real-world news queries. These survey signals should guide deployment safeguards, monitoring, and research into how generative systems integrate with journalistic standards and verification workflows.