AI chatbots may worsen symptoms in first-episode psychosis patients

A new observational study published in JAMA Psychiatry suggests that conversational AI chatbots may worsen symptoms in vulnerable psychiatric patients, especially those experiencing a first episode of psychosis. The findings are rare but clinically significant, adding to concerns that rapid adoption of chatbots in consumer and clinical settings may create risks not yet fully understood. Researchers led by Zachary Bergson at Vanderbilt University Medical Center reviewed electronic health records for approximately 216,000 behavioral health patients treated between December 2022 and April 2026. The team used keyword searches and manual screening to identify patients with documented psychotic disorders and references to AI. That process yielded 73 individuals with relevant records. Reviewers then classified cases into groups based on whether chatbot use appeared to worsen symptoms, had a neutral effect, or involved delusional beliefs about AI without documented chatbot interaction. Among the 73 patients, 28 had records indicating that chatbot use worsened psychosis. Of those, about 61% were experiencing their first psychotic episode, a notably higher share than in the other categories. The authors said the early phase of psychotic illness may make individuals particularly vulnerable to AI interactions that intensify symptoms. The study also categorized how chatbots may have contributed to clinical worsening. In 64% of the worsening cases, chatbots acted as amplifiers, validating distorted or delusional ideas. In 21%, the AI itself became the object of delusional beliefs. Another 11% were classified as catalysts, in which chatbot interaction appeared to trigger new symptoms in patients who had previously been healthy. No cases fit the co-author category, in which users and chatbots gradually build harmful delusional narratives through continued conversation, and one patient did not fit any group. The study does not claim that chatbots cause psychosis. Instead, it highlights the importance of assessing AI use during psychiatric care, particularly for patients in early stages of illness or those showing signs of psychosis, delusion, or impaired reality testing. The researchers argue that clinicians should routinely ask about chatbot use as mental-health-related AI tools become more widespread.