Self-Reported Survey on the Use of Artificial Intelligence Chatbots by Adult Patients on an Inpatient Psychiatric Unit

OBJECTIVE: Despite widespread concern about the use of artificial intelligence (AI) chatbots (e.g., ChatGPT, Claude, and Google Gemini), especially among individuals with mental health problems, little to no empirical data about such use have been obtained from those experiencing severe mental illness. The authors aimed to characterize the prevalence and perceptions of chatbot use among adults being discharged from an inpatient psychiatric unit. METHODS: Adults (N=100) pending discharge from an inpatient unit were surveyed about their chatbot use, including general (i.e., non-mental health) use, use specifically for mental health purposes, and perceived helpfulness or harms of and barriers to use. RESULTS: Overall, 61% of the participants reported using an AI chatbot, with 30% using such chatbots specifically for mental health purposes. Chatbot use was most common among patients ages <39 years, who accounted for 87% of the chatbot users and for 90% of those who had used a chatbot for mental health purposes. Among those who had used a chatbot for mental health purposes, most (80%) rated the chatbots positively, perceiving these tools as nonjudgmental and accessible, and 37% had used a chatbot during a psychiatric crisis. Disclosure of this use to care providers was uncommon, with only 13% of those who used chatbots for mental health purposes reporting openly discussing such use with their providers, and 37% actively intending not to disclose such use. CONCLUSIONS: AI chatbot use among individuals hospitalized for psychiatric care was found to be common, clinically relevant, and largely undisclosed, highlighting the need for clinicians to ask their patients about chatbot use.