More Americans got news from chatbots, but few preferred them

Pew’s October 19, 2026 fact sheet found that 15% of U.S. adults got news from AI chatbots at least sometimes, up from 9% in 2025. Only 1% preferred chatbots as their news platform. The two questions measure different things: reporting that they get news from a service and naming it as the platform they prefer. ^[1]

The 2026 findings came from 5,166 adults surveyed August 24–30 through the American Trends Panel, a randomly recruited national panel. Interviews were conducted online and by telephone, and the results were weighted to represent U.S. adults. This measures reported habits, rather than independently tracking people’s browsing or checking chatbot answers. ^{[1] [3]}

News websites or apps reached 66% at least sometimes, compared with 59% for social media and 70% for search. Because respondents could use several routes, these rates describe overlapping reported use. ^[1]

In the 2025 survey, the perceived-accuracy question went to 1,240 respondents who reported getting chatbot news at least rarely, including rare users as well as often or sometimes users. Of that group, 49% reported coming across news they thought inaccurate at least sometimes; 16% said this happened often or extremely often and 29% were unsure. These are users’ judgments from August 18–24, 2025, rather than researchers grading answers. ^{[2] [4]}

That base is broader than the at-least-sometimes measure behind the 9% and 15% use figures. It also predates the summer 2026 survey. The findings should not be combined as if they measured the same audience or question. Popularity does not establish whether an individual chatbot answer is correct, and these surveys do not measure whether publishers lost referral traffic. ^{[1] [4]}

Television led the 2026 preference question at 28%, followed by websites or apps at 21% and social media at 19%. Those figures describe respondents’ preferred platform, alongside the separate measure of all the platforms they reported getting news from. ^[1]

About this item: Retrospective report. No first-hand reporting or independent testing is claimed. Corrections policy.