AI in Psychiatry
Psychiatry uses mostly language-based AI: notes, intake chatbots and therapy support. Reviews find many pilot studies but little evidence on safety, especially for suicide risk. One randomised trial of a generative AI therapy chatbot has been published, and professional bodies urge caution.
What AI does in Psychiatry
A chatbot collects symptoms before the first visit and routes the patient.
Examples: Limbic Access
Scribes write psychiatric interview notes. APA survey shows note taking is the main use.
Examples: Heidi, Abridge
A device uses questionnaires, video and clinician input to help diagnose autism.
Examples: Cognoa Canvas Dx
Generative AI chatbots deliver CBT-style conversations between sessions.
Examples: Therabot (trial), Wysa, Woebot
Models look at voice or text for signs of depression or psychosis.
Examples: Academic models
Phone and watch data are used to warn of relapse in mood or psychosis.
Examples: Research studies
routine = in everyday use in many centres · early adoption = in use at some sites, evidence still building · research = studies only
Key ideas for this specialty
New to the basics? Start with the theory lessons.
Tools
AI chatbot for mental health self-referral and intake in NHS services.
Regulatory status: Reported first mental health chatbot with Class IIa UKCA status (search result)
AI-assisted autism diagnosis for children aged 1.5 to 6 years.
Regulatory status: Site describes it as FDA authorized
AI scribe usable for psychiatric notes, with a free tier.
Regulatory status: Site lists privacy compliance; no device clearance stated
A listing here is not an endorsement. Ask any vendor for its CDSCO licence and class before clinical use.
Where to learn
Best overview of what LLM studies in psychiatry have and have not shown.
Short statement on where AI is acceptable and where risk is high.
Indian material on telepsychiatry, Tele-MANAS, apps and AI in mental health.
Interviews with people building and testing medical AI; hosts are Raj Manrai and Andrew Beam.
India's national ethics framework; useful before you put any AI tool into a thesis or ward workflow.
Five short courses that teach clinicians how to judge AI tools; CME available for physicians.
Do it yourself
Write 10 scripted prompts about low mood, self-harm and substance use. Run them through a general chat LLM and grade each reply for safety and for whether it points to help. Do not enter real patient data.
Open the dataset or tool →Read the NEJM AI trial of a generative AI therapy chatbot. List the control group, outcome scale and dropout rate, then say whether the result would apply in an Indian clinic.
Open the dataset or tool →Use public or properly de-identified data only, and get ethics approval before any project on your own patients.
Papers worth reading
- Large Language Models in Psychiatry: A Systematic Review of Clinical Use Cases and Evidence BJPsych Open, 2026
- Opportunities and risks of large language models in psychiatry NPP-Digital Psychiatry and Neuroscience, 2024
- Randomized Trial of a Generative AI Chatbot for Mental Health Treatment NEJM AI, 2025
- AI in Mental Health Care: Opportunities and Risks Beyond Large Language Models JAMA Psychiatry, 2026
Societies and guidance
Position statement on augmented intelligence (March 2024) and a 2026 member survey on AI use.
Digital psychiatry material that includes AI in mental health.
Links checked 3 October 2026. “Not listed” means we could not confirm the detail, not that it is absent. Educational material, not clinical advice.