AI Scribes for Group Therapy Sessions

Why group is the hardest test for an AI scribe

Group sessions violate almost every assumption in a two-speaker medical-visit model. There are 4–12 speakers instead of two. Turns overlap. Members interrupt each other. Some sit far from the microphone. And the deliverable is not one note but N per-member notes plus an overall group summary.

All of that stresses the diarization layer, the LLM's ability to keep speakers separate, and the vendor's product design around per-member workflows.

What to test in a group pilot

Run a real, unedited group with your candidate scribe and grade it against your normal note process.

1. Speaker separation accuracy across all attendees (not just the loudest three).
2. Per-member note quality — is each note distinct, or is there cross-contamination?
3. GIRP output aligned to each member's individual treatment plan.
4. Handling of the therapist prompts and the group process (not just the individual sharing).
5. Time-to-final-notes vs. your current baseline.

Our 2026 group-therapy picks

Twofold Health is our Editor's Pick for outpatient group work — per-member GIRP output, treatment-plan-aligned goals, and audit-ready risk documentation on every session. Blueprint and Eleos remain strong alternatives, particularly Eleos for enterprise SUD group settings. Mentalyc is capable and price-competitive.

Freed, Heidi, Nabla, and Suki are not designed for groups and should be avoided for group documentation regardless of how well they perform in one-on-one visits.

Modality-specific considerations

Different group modalities stress different parts of the pipeline.

• DBT skills groups: high therapist-teaching share; scribes over-attribute content to the therapist unless prompted otherwise.
• IOP process groups: high emotional intensity, frequent tears; ASR quality drops noticeably.
• SUD education groups: heavy jargon (MOUD, cravings, relapse prevention); domain-tuned scribes clearly outperform.
• Family/couples: fewer speakers but relational dynamics — the note must capture the dyad, not just individual content.

Billing alignment: CPT 90853 and per-member documentation

CPT 90853 (group psychotherapy) requires per-member documentation. Payers audit for evidence that each billed member received clinically meaningful intervention — a single group summary is not sufficient. Any scribe you use for groups must produce distinct, member-specific notes; check this in the pilot output before generalizing.