Use of Generic Large Language Model Chatbots for Mental Health Issues Among Patients Seeking Psychotherapy: Cross-Sectional Mixed Methods Analysis in a German University Hospital Outpatient Clinic

Key Findings

• — Of 812 outpatients who reported on chatbot use, 245 (30.1%) had used LLM chatbots for mental health; among 217 with frequency data, 31.8% once, 54.4% weekly and 13.8% daily
• — 61.4% (121/197) had used a chatbot in a self-reported personal crisis situation, and 67.7% (82/121) of those found the experience unhelpful
• — Users were more likely to be younger, female, better educated, students or unemployed
• — Use was higher with anxiety disorder diagnoses (persisting after adjustment for age) and lower with somatoform disorders
• — The authors suggest asking about chatbot use in psychosocial assessment, especially crisis use, symptom-reinforcing use and replacement of human support

Methodology Notes

Consecutive outpatient sample over 6 months (calendar dates not stated in the abstract); quantitative questionnaire plus an open-ended question analysed qualitatively; descriptive and exploratory inferential statistics. Self-report; denominators vary by item (217, 197, 121). Received 2026-06-11, accepted 2026-09-08, published 2026-10-09 (JMIR AI vol 5, e104297). Verification: PubMed 42855158 efetch (structured abstract, affiliations, history); Crossref published 2026-10-09; ai.jmir.org direct 202. Full text not read.

Sources