AI chatbots have quietly become a news source. Ask one what happened today, and it will summarize, condense, and explain — no scrolling, no tabs, no paywalls. Plenty of people find that genuinely useful.
But according to MIT Technology Review, a study led by Pattie Maes and her colleagues at the MIT Media Lab points to a significant problem with relying on chatbots this way. In the study, participants evaluated paired news headlines and images over the course of four weeks. MIT Technology Review reports that participants were initially 21% more — the source excerpt available here cuts off before completing that finding, so the full outcome of the comparison isn't captured in what's been published to this brief.
What is clear is the shape of the concern: the researchers are examining not just whether chatbots deliver accurate information, but what happens to the person on the other end over time. The four-week design matters. A single chatbot summary might be perfectly fine. The question Maes and her team are probing is what repeated exposure does to how people judge what they're reading — whether the convenience of having a machine pre-digest the news changes readers' own evaluation of it.
That framing puts this study in a different category from most AI-and-misinformation research, which tends to focus on the output: is the model right, is it hallucinating, is it biased. This one focuses on the reader.
It matters because hundreds of millions of people are now forming habits around AI assistants before anyone has measured what those habits do to our judgment — and habits, once formed, are much harder to study than they are to change.