A 2024 study conducted by researchers from MIT and the University of California-Irvine found that individuals who recounted an event through interactions with a generative chatbot developed more than three times as many false memories compared to a control group. The study also indicated that participants maintained confidence in these false memories for a week following the chatbot interactions. Psychologist Elizabeth Loftus, PhD, co-authored the 2024 study.

In 2026, a KFF poll revealed that 32% of U.S. adults utilized AI chatbots for health information over the previous year. The poll found that many of these adults who inquired about physical or mental health with AI chatbots did not subsequently consult with a clinician.

Also in 2026, a study published in Nature Medicine observed that members of the public using large language models did not perform better than control groups in identifying relevant medical conditions or selecting appropriate courses of action. That study also found that large language models demonstrated superior performance when tested independently, compared to when they were used by members of the public.