📄 health informatics

Generative Artificial Intelligence in Psychotherapy Practice: A Global Online Survey of Mental Health Professionals' Adoption

A global survey of 766 mental health professionals reveals that over half are already using generative AI tools, primarily for treatment planning and administrative tasks, yet this widespread adoption occurs despite a significant lack of institutional guidance, formal training, and regulatory frameworks.

Original authors: Blease, C., Hagström, J., Gaab, J., Carey, A., Cipriani, F., Gorman, C., Nascimento, A. F., Fitzgerald, A., Holtz, L., Schwarz, J., Strudwick, G., Tibbs, M., Torous, J., Cornelius-White, J. H. D.

Published 2026-06-22

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Blease, C., Hagström, J., Gaab, J., Carey, A., Cipriani, F., Gorman, C., Nascimento, A. F., Fitzgerald, A., Holtz, L., Schwarz, J., Strudwick, G., Tibbs, M., Torous, J., Cornelius-White, J. H. D.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine the world of mental health therapy as a vast, busy kitchen where chefs (therapists) are preparing special meals (treatment) for hungry guests (clients). For years, these chefs have been doing everything by hand: chopping vegetables, writing down recipes, and cleaning the counters.

Now, a new, powerful robot assistant called Generative AI (like ChatGPT) has entered the kitchen. This paper is a global survey asking 766 of these chefs: "Are you using this robot? What are you using it for? And is your boss telling you how to use it?"

Here is what the study found, explained simply:

1. The Robot is Already in the Kitchen

The biggest surprise? More than half (54.6%) of the chefs are already using the robot.
It's not just a few early adopters; it's widespread. The most popular robot is ChatGPT, used by almost everyone who is using any AI at all. It's like if 8 out of 10 people using a robot in the kitchen were using the exact same brand.

2. What Are They Asking the Robot to Do?

The chefs aren't letting the robot cook the main meal (the therapy session itself) yet. Instead, they are using it for the "prep work" and "cleaning up."

Planning the Menu: The most common use is helping to plan the treatment (42%).
Cleaning the Counters: Almost as common is using it for administrative paperwork (41%).
Writing Handouts: They also use it to create educational materials for clients (40%).
The Result: About 83% of the chefs who use the robot say it makes their workload lighter. It's like having a dishwasher that actually works; they feel less tired.

3. The "Silent Kitchen" Problem (No Rules, No Teachers)

Here is the tricky part. While the chefs are using the robot, nobody has really taught them how.

No Training: Over 80% of the chefs said they have never received any professional training on how to use these AI tools. They are learning by trial and error.
No Bosses Speaking Up: Only about 18% said their employer or professional organization encouraged them to use the robot. Most bosses are staying silent, and a few have even said "No robots allowed."
The Analogy: Imagine a kitchen where half the staff is using a high-tech robot to chop onions, but the head chef has never given them a manual, hasn't told them if the robot is safe, and hasn't taught them what to do if the robot starts chopping the wrong thing.

4. Who is Using the Robot?

The study found that younger chefs are much more likely to use the robot than older ones. Also, chefs working in rural areas (small towns) are using the robot more than those in big cities. It's as if the chefs in the countryside are more eager to try the new gadget to help them get things done.

5. The Big Takeaway

The paper concludes that the robot is running faster than the rulebook.
The chefs are adopting this technology quickly because it helps them with paperwork and planning, but the "safety manuals" (training, rules, and ethics) are lagging far behind. The authors argue that before we let the robot take over more tasks, we need to write the rulebook, teach the chefs how to use it safely, and make sure the robot doesn't accidentally serve a bad meal.

In short: Mental health professionals are already using AI to help with their daily work, and they like it because it saves time. However, they are mostly doing this on their own, without official training or clear rules from their organizations. The technology is moving faster than the safety guidelines.

Technical Summary: Generative Artificial Intelligence in Psychotherapy Practice

Problem Statement
Generative artificial intelligence (GenAI) tools, particularly large language models (LLMs) like ChatGPT, Google Gemini, and Microsoft Copilot, are being rapidly adopted across healthcare. While these tools offer potential for alleviating administrative burdens and enhancing clinical workflows, empirical data regarding their actual adoption patterns, specific use cases, and the institutional context within global mental health practice remains scarce. Prior research has largely focused on physicians in high-income countries or on attitudes rather than behaviors. There is a critical gap in understanding how mental health professionals specifically are integrating GenAI into routine psychotherapy, including the prevalence of use, the tools selected, the perceived impact on workload, and the extent of institutional guidance or training.

Methodology
This study employed a pre-registered, cross-sectional mixed-methods survey design conducted between January and March 2026.

Participants: A global convenience sample of 766 licensed mental health professionals (psychotherapists, psychologists, counselors, nurses, and psychiatrists) from 30 countries who provide psychotherapy. Recruitment occurred via professional networks and email lists, deliberately avoiding social media to prevent bot contamination.
Instrument: An online survey administered via LimeSurvey, consisting of four sections. This paper reports on Section A (AI use behaviors) and Section D (demographics). The survey captured data on purposeful GenAI use, specific tools employed, clinical and administrative purposes, perceived workload impact, and institutional context (encouragement, prohibition, training).
Analysis: Descriptive statistics were calculated for all variables. Bivariable associations between AI use and predictors (age, gender, role, setting, etc.) were examined using chi-square, Fisher's exact, t-tests, or Mann-Whitney U tests. Variables with $p < .10$ in bivariable analyses were entered into a multivariable binary logistic regression model to identify independent predictors of adoption.

Key Results

Prevalence of Use: 54.6% ( $n=418$ ) of respondents reported purposely using at least one GenAI tool in their psychotherapy practice.
Tools: ChatGPT was the dominant tool (84.7% of users), followed by Google Bard/Gemini (17.7%) and Microsoft Copilot/Bing AI (12.7%).
Purposes: The most common uses were assisting with treatment planning (41.9%), managing administrative tasks (41.4%), and generating psychoeducational materials for clients (39.7%). Other uses included session note documentation (30.6%) and supporting reflective practice (32.5%).
Workload Impact: Among users, 82.8% reported that GenAI tools reduced their overall work burden.
Institutional Context: Institutional support is lagging significantly. Only 18.1% of respondents reported employer encouragement to use AI, while 8.2% reported prohibitions. Crucially, 81.1% of all respondents had received no professional training on AI use.
Predictors: Multivariable logistic regression identified younger age and rural practice settings as significant independent predictors of GenAI adoption. Professionals in urban and suburban settings had significantly lower odds of use compared to those in rural settings. Professional role, gender, and country income group were not significant predictors in the final model.

Significance and Claims
The authors claim that this study provides the first global characterization of GenAI adoption across a diverse, multi-professional sample of mental health clinicians. The findings indicate that adoption is already substantial and widespread, concentrated on both administrative relief and core clinical support tasks like treatment planning.

The paper emphasizes a critical misalignment: adoption rates are high and growing, yet formal training, institutional guidance, and regulatory oversight substantially lag behind. The authors argue that this "governance failure" poses ethical risks, as clinicians may be utilizing tools (often general-purpose, consumer-grade LLMs) without the competency to detect limitations such as hallucinations or data privacy issues.

The study concludes that there is an urgent need for:

Evidence-based competency frameworks for safe and ethical AI use in psychotherapy.
Clear regulatory and governance structures.
Professional education to support the integration of AI, ensuring it enhances rather than compromises the therapeutic alliance and clinical accuracy.

The authors explicitly note that their findings are based on a convenience sample, which may overrepresent those with an existing interest in digital health, and that the study is limited by language barriers (English/German) and a lack of low-income country representation. They call for future research to examine clinical outcomes, longitudinal adoption trajectories, and the specific dynamics of AI use in low- and middle-income contexts.