Generative AI Adoption in Psychotherapy Practice Among Mental Health Professionals: Multicountry Cross-Sectional Survey.

Background

Generative AI (GenAI) tools, such as OpenAI's ChatGPT, Google's Gemini, and Microsoft Copilot, are being adopted across health care settings. Despite the popularity of GenAI, empirical data on the extent of adoption by mental health clinicians in routine psychotherapy practice are scarce.

Objective

This study aimed to characterize purposeful GenAI use among mental health professionals providing psychotherapy, including tools used, purposes, perceived workload effects, and the organizational context surrounding use.

Methods

We conducted a cross-sectional online survey between January 13, 2026, and March 30, 2026. A convenience sample of trained mental health professionals who had provided psychotherapy within the previous 12 months was recruited through professional associations and academic and clinical mailing lists, with snowball dissemination by email. The survey assessed lifetime purposeful GenAI use in psychotherapy practice; tools and purposes of use; perceived work burden reduction; and, for the previous 12 months, employer or professional organization encouragement, prohibition, and participation in AI training. Descriptive statistics were calculated. Univariable logistic regression models were used to examine associations between GenAI use and candidate factors. Variables meeting an entry criterion of P<.10 were entered into a multivariable logistic regression; statistical significance was set at P<.05.

Results

A total of 766 mental health professionals who provide psychotherapy in 30 countries completed the survey. Of these, 54.6% (n=418, 95% CI 51.0%-58.1%) reported having purposely used at least 1 GenAI tool in clinical psychotherapy practice. ChatGPT was the most frequently used tool (354/418, 84.7%, 95% CI 81.2%-88.2%). The most commonly reported clinical purpose was assisting with treatment planning (175/418, 41.9%, 95% CI 37.1%-46.7%), followed by managing administrative tasks (n=173, 41.4%, 95% CI 36.6%-46.1%) and generating psychoeducational materials for clients (n=166, 39.7%, 95% CI 35.0%-44.4%). Among AI users, 82.8% (346/418, 95% CI 79.1%-86.4%) reported that these tools reduced their overall work burden. Only 18.1% (139/766, 95% CI 15.4%-20.9%) of respondents reported institutional encouragement to use GenAI tools, while 81.1% (n=621, 95% CI 78.3%-83.9%) reported not having received any professional training on AI use. In multivariable logistic regression (α=.05), younger age was associated with GenAI use, while practice setting showed mixed evidence of association. The overall adjusted effect of practice setting did not reach statistical significance (P=.07), although urban practitioners had lower odds of GenAI use than rural practitioners (adjusted odds ratio 0.614, 95% CI 0.388-0.971; P=.04).

Conclusions

This study extends previous predominantly national and profession-specific research by providing a multicountry, multiprofessional snapshot of reported GenAI use in psychotherapy practice. Use extended beyond administrative tasks into clinically consequential activities, including treatment planning and psychoeducation, while participation in professional AI training remained relatively uncommon. Findings highlight the need for professional bodies, health systems, and training programs to develop practical guidance and competencies for safe, privacy-conscious, and clinically appropriate GenAI use.