Chatbot Use May Worsen Psychosis Symptoms in First-Episode Patients, Study Finds

A new study from Vanderbilt University Medical Center suggests that interacting with chatbots may worsen psychiatric symptoms in vulnerable individuals, particularly those experiencing a first episode of psychosis. The findings were published in JAMA Psychiatry.

Researchers reviewed electronic health records of about 216,000 behavioral health patients from December 2022 to April 2026. After keyword searches and manual screening, they identified 73 patients with a psychosis diagnosis and documented AI-related content. Reviewers classified cases into three groups: symptom worsening linked to chatbot use, neutral impact, and delusions about AI. In 28 patients, chatbot use appeared to aggravate psychotic symptoms, and about 61% of these cases were first-episode psychosis, a significantly higher share than in the other groups.

Among the worsening cases, 64% were categorized as “symptom amplifiers,” in which chatbots reinforced distorted beliefs; 21% as “delusion objects,” where AI itself became the focus of delusions; and about 11% as “catalysts,” where new symptoms emerged during interaction. No cases involved co-constructing harmful fantasy narratives, and one patient did not fit any category. The study does not claim chatbots cause psychosis, but it highlights potential risks for susceptible people and suggests that clinicians should regularly assess patients’ AI use during psychiatric care.