Quick orientation: Mental health therapy involves a specific combination of factors that makes AI tool adoption more fraught than in most other professions: highly sensitive personal information, a therapeutic relationship that depends on trust and confidentiality, and professional ethics standards that require careful evaluation of any tool that handles client information. This guide is more cautious than most in this category — deliberately so.
Why Caution Is the Right Default for Therapy AI Tools
Client confidentiality in psychotherapy is not just a HIPAA obligation — it is foundational to the therapeutic relationship. Clients disclose things in therapy they have disclosed nowhere else, with the explicit understanding that the information is protected. Any AI tool that processes therapy session content, client notes, or session recordings handles information of extraordinary sensitivity, and the consequences of data breaches or inappropriate data use are particularly severe in this context.
Beyond confidentiality, several state licensing boards have issued guidance on AI use in clinical practice that individual therapists should review before adopting any tool that touches client information. Ethics standards vary by professional association (APA, NASW, AAMFT, ACA) and are evolving in response to AI adoption.
Where AI Can Help Therapists
Session note drafting — with strictly de-identified content — is the clearest appropriate use case. Some therapists use AI tools to structure their post-session notes by describing the session in general terms (presenting concern, themes discussed, clinical observations, treatment plan updates) without patient-identifying information. The AI structures the note; the therapist adds the identifying information and clinical content in their secure EHR system.
Administrative documentation — intake forms, consent documents, treatment plan templates, progress note structure — involves lower sensitivity than session content and is an appropriate area for AI assistance with careful review.
Continuing education and clinical research — using AI to summarize research literature on specific therapeutic approaches, synthesize evidence on treatment modalities, or prepare case conceptualization frameworks — is a low-risk, high-value use case that does not involve client information at all.
Where AI Creates Serious Risk for Therapists
Recording or transcribing actual therapy sessions using AI tools is the highest-risk application for therapists. Even with a HIPAA-compliant tool and a BAA in place, recording therapy sessions with AI raises client consent requirements, increases the attack surface for data breaches involving highly sensitive disclosures, and may violate state recording consent laws or professional ethics standards. Most professional associations' ethics guidance on this is either silent (because it is developing) or cautionary.
AI chatbots or tools positioned as therapeutic supplements for clients between sessions carry a different kind of risk — the potential for clients to disclose crisis-level information to an AI tool that does not have appropriate escalation pathways. Therapists recommending AI tools to clients for between-session support carry responsibility for evaluating whether those tools are appropriate for their specific clients' presentations.
Tools to Consider Carefully
Specialized therapy note platforms (e.g., Mentalyc, Blueprint)
Evaluate carefully before use with client contentSeveral platforms built specifically for mental health documentation offer HIPAA-compliant session note assistance. Evaluate each tool's data handling practices, BAA availability, and whether your state licensing board has issued guidance on AI documentation tools before adoption. No specific recommendation here — the evaluation depends on your state's requirements and professional association guidance.
ChatGPT (de-identified administrative use only)
Administrative tasks only — never with client informationFor drafting intake document templates, scheduling communications, and continuing education research with fully de-identified content. Never use general-purpose AI tools with client-identifying information or session content.
The Professional Ethics Standard That Applies
Before adopting any AI tool in a therapy practice, the relevant question is not "does this tool save me time?" but "have I satisfied my ethical obligation to ensure this tool handles client information appropriately?" That question requires more than accepting a vendor's HIPAA compliance claim — it requires understanding what the tool actually does with session data, how long it is retained, whether it is used for model training, and what your state licensing board and professional association say about tools in this category.