MIT Study Finds Over-Reliance on Chatbots Reduces Independent Misinformation Detection

A Massachusetts Institute of Technology study released in April tracked 67 participants over four weeks to measure how AI assistance affects the ability to identify misinformation in news headlines and images. Participants answered questions about real and manipulated content both with and without an AI assistant built on GPT-4o and integrated with Google search.

The chatbot provided hints such as directing attention to visual details like a police badge that indicated an image was fabricated. When participants used the AI tool, their accuracy rose by 21 percent. Without the tool in the fourth week, their performance fell 15.3 percent compared with earlier unassisted results.

The study found that AI systems optimized for quick, definitive answers produced different outcomes than systems that asked guiding questions. Participants who received prescriptive responses often followed the chatbot’s conclusion without further scrutiny. About one-quarter of participants believed their detection skills were improving even as measured performance declined.

The authors noted that most participants were from the US and UK and that longer studies would be needed to determine whether the observed decline continues beyond four weeks. Earlier research cited in the study includes a 2025 Lancet paper on AI diagnostic tools and a neuroscientist’s warning about cognitive effects of heavy AI use.

"When we’re interacting with AI, we feel we’re becoming better at certain tasks and there’s enough research that shows we are not," said Anku Rani, a PhD student at MIT and co-lead author of the study.