Potentially Harmful Consequences of Artificial Intelligence (AI) Chatbot Use Among Patients With Mental Illness: Early Data From a Large Psychiatric Service System

Key Findings

• — 53,974 patients had at least one contact in the period and 10,712,856 notes were entered; 181 notes from 126 unique patients contained one of the 22 search terms, at a rate that increased over time
• — Notes from 38 unique patients (39% female, median age 28) were judged compatible with potentially harmful consequences of AI chatbot use on mental health
• — Dominant psychopathology among the 38: delusions (n=11), suicidality or self-harm (n=6), feeding or eating disorder (n=5); mania or hypomania, obsessions or compulsions, depression and anxiety each fewer than five
• — Described patterns included the chatbot as object or consolidator of delusions, reinforcement of hypomania, compulsive use to relieve obsessions, enabling calorie restriction or counting, and information-seeking on suicide methods
• — Notes also described 32 patients using chatbots for seemingly constructive mental-health purposes (psychoeducation, talk therapy, companionship, diagnostics) and 20 for practical tasks
• — The authors state the notes are not evidence of causation, patients were not systematically asked about chatbot use, the search was narrow, and the results do not speak to incidence

Methodology Notes

Electronic-health-record keyword search with 22 terms (chatbot, ChatGPT and ten alternative spellings of each); independent assessment by two authors with consensus on disagreements; counts below five suppressed for identifiability. Approved by the Central Denmark Region with a consent waiver; register-based studies are exempt from ethics-board review in Denmark. No funding reported. Published online 6 February 2026 (print issue April 2026), CC BY-NC-ND. The Wiley page blocks automated access; verified from the Europe PMC full-text deposit PMC12967755, which appeared between 15 and 17 September 2026.

Sources