Guidelines needed as AI embraced by mental health practices
Artificial intelligence (AI) tools are overwhelmingly being used in mental health practices to support administrative tasks, but University of Queensland researchers have found they potentially lack privacy, accuracy and governance safeguards.
AI tools are increasingly being employed in mental health practices to assist with administrative duties, but University of Queensland researchers warn they may lack privacy, accuracy and governance safeguards. A survey of 278 clinicians, alongside 12 in-depth interviews, revealed that nearly 80% of mental health clinicians use AI to transcribe and listen to client sessions.
Many also rely on AI for administrative tasks, though it proved less reliable in complex clinical situations. Despite weak industry guidelines, AI has already been embraced by many practices, according to Ph.D. candidate Benjamin Johnson from UQ's National Centre for Youth Substance Use Research (NCYSUR). Johnson notes that AI complements clinicians well, particularly for routine administrative work, but is not yet capable of replacing clinicians in complex cases requiring accurate diagnoses and treatment recommendations.
The study found that 43.5% of surveyed clinicians used AI daily for at least one administrative or clinician-support task, such as drafting reports, writing case notes, conducting research and transcribing sessions. Most clinicians found AI to be superior to humans for tasks like preparing documents, research and note-taking, allowing them more time with clients, being more present during sessions and making clinical decisions.
Mental health work can be demanding and emotionally exhausting, so many saw AI as a way to alleviate cognitive pressures during administrative tasks. A separate study by UQ researchers reviewed 66 studies evaluating large language models like ChatGPT in mental health settings. Co-author Associate Professor Janni Leung emphasized that most clinicians agreed AI is not a replacement for therapists, their clinical judgment or therapeutic relationships.
Simulated clinical scenarios in many studies resulted in AI underperforming compared to humans, highlighting the importance of human therapeutic relationships. Concerns about privacy, data security, accuracy and client consent were raised by surveyed clinicians. Around 22% of mental health workplaces banned AI during client sessions, while 5% strictly prohibited AI for administrative or documentation tasks.
Conversely, 31% of workplaces encouraged appropriate AI use for administrative tasks, and 30% permitted its use during client sessions. As AI adoption outpaces governance, clinicians called for practical guidance rather than blanket bans. Johnson stated that people are using AI regardless of organizational policies, so clear, industry-wide standards could enhance client safety, privacy protections and clinician confidence.
The findings are published in the Journal of Medical Internet Research, and a systematic review on how large language AI models perform in mental health settings appears in JMIR AI.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
