Not All Clinical Notes Are Created Equal
When people think of AI medical scribing, they typically imagine a primary care visit: patient presents with a cough, physician examines, diagnoses upper respiratory infection, prescribes rest and fluids. The documentation is relatively straightforward — history of present illness, review of systems, assessment, and plan.
Mental health documentation is fundamentally different. A therapy session is a conversation — often an emotionally charged one — where the therapeutic relationship itself is the intervention. The content is sensitive in ways that go beyond standard PHI. Patients disclose trauma, relationship conflicts, substance use, suicidal ideation, and deeply personal struggles. Documenting these sessions requires a level of nuance and discretion that most AI systems are not designed to handle.
The Unique Challenges
Mental health documentation has several characteristics that set it apart from other specialties.
First, there is the distinction between psychotherapy notes and the medical record. Under HIPAA, psychotherapy notes — the therapist's personal observations and analysis recorded during a session — receive special protection. They are stored separately from the medical record and cannot be disclosed without specific patient authorization, even to other treating providers. An AI documentation tool must understand this distinction and never include psychotherapy note content in the standard progress note.
Second, progress note formats in mental health vary significantly by modality and setting. A psychiatrist prescribing medication may use a standard SOAP format. A therapist conducting CBT might document using a session-by-session treatment plan update. A group therapy facilitator needs to document individual participation within a group context. A one-size-fits-all documentation template fails all of these use cases.
Third, DSM-5 coding adds complexity. Mental health diagnoses often involve specifiers (mild, moderate, severe, in partial remission, in full remission), comorbidities that interact in clinically significant ways, and diagnostic changes over the course of treatment. An AI system that simply pattern-matches keywords to diagnosis codes will produce inaccurate and potentially harmful documentation.
Why General-Purpose AI Scribes Fall Short
Most AI medical scribes were built for high-volume, procedure-oriented specialties — primary care, urgent care, orthopedics. They are optimized for encounters with clear chief complaints, physical exam findings, and discrete diagnoses. Mental health encounters rarely follow this pattern.
A therapy session might spend 45 minutes exploring a patient's relationship with their mother. The clinically relevant documentation from that session might be three sentences about the patient's attachment patterns, their emotional regulation progress, and a treatment plan adjustment. A general-purpose AI scribe would either generate a verbose, inappropriately detailed transcript of the conversation (violating the spirit of psychotherapy note protections) or produce a note so generic as to be useless.
The challenge is not transcription accuracy — it is clinical judgment about what to include, what to exclude, and how to frame therapeutic observations in a way that is accurate, useful for continuity of care, and appropriate for the medical record.
How WhisperFlow Approaches Mental Health Documentation
We spent six months working with psychiatrists, psychologists, licensed clinical social workers, and marriage and family therapists before building our mental health documentation module. What we learned shaped every design decision.
The system recognizes the type of mental health encounter (medication management, individual therapy, group therapy, psychological assessment, crisis intervention) and applies the appropriate documentation framework. For medication management visits, it generates a focused note covering current symptoms, medication response, side effects, and prescription changes. For therapy sessions, it produces a concise progress note that captures therapeutic themes, interventions used, patient response, and treatment plan updates — without reproducing the session content verbatim.
Critically, the system is designed to exclude by default. Rather than capturing everything and letting the provider delete, it captures only what belongs in a progress note and flags anything that might warrant inclusion in separate psychotherapy notes for the provider's private records. This approach respects both HIPAA requirements and the therapeutic relationship.
DSM-5 coding is handled with full specifier support. When the encounter discussion indicates a change in symptom severity or remission status, the AI suggests updated diagnostic codes with specifiers rather than simply carrying forward the existing code. This improves diagnostic accuracy and supports appropriate treatment planning.
The Therapeutic Alliance Factor
One thing we heard repeatedly from mental health providers during development was concern about the therapeutic alliance. Would patients be comfortable knowing an AI was listening to their therapy session? Would it change what they were willing to disclose?
This is a legitimate concern, and the answer depends on implementation. WhisperFlow provides customizable patient consent workflows specifically for mental health settings, including the ability to pause or stop recording at any point during the session. Providers can also choose to use the tool only for medication management visits and not for therapy sessions — the system supports partial adoption.
Early feedback from providers using WhisperFlow for mental health documentation has been encouraging. Several therapists report that the reduced documentation burden actually improves the therapeutic relationship because they are more present during sessions — not mentally cataloging what they need to document later. One psychologist told us: "I used to spend the last five minutes of every session half-listening while I mentally composed my note. Now I am fully present until the session ends."
Moving Forward Thoughtfully
AI documentation in mental health requires more care, more nuance, and more humility than in other specialties. We do not claim to have solved every challenge. But we believe that mental health providers deserve the same documentation relief that their colleagues in primary care and surgery are beginning to experience — they just need a tool that understands their unique requirements.
That is what we built, and we continue to refine it in close partnership with the clinicians who use it every day.