The short version Group therapy is the hardest ambient-capture workload in behavioral health: 4–12 speakers, overlapping talk, per-member progress notes, and CPT 90853 billing that requires documentation for each participant. Most therapy-first scribes can *record* a group session; far fewer produce clean per-member notes.
**Our current recommendation for solo and small-group US practices running standard process groups is Twofold — its group workflow generates a shared session summary plus per-member progress entries that map cleanly to the Golden Thread from each member's treatment plan, and it lands at $19 for the first month. For established practices running EMDR groups, DBT skills groups, or IOP-style curricula, Mentalyc remains the deeper option on template breadth ($89.99/mo Group/Couples tier). Upheal** is the pick when you need EU data residency or multilingual groups.
What "group support" actually means Vendors use the phrase loosely. Ask specifically:
1. Speaker separation — can it label speakers (Member A, Member B) reliably enough that a per-member note isn't fabricated? 2. Per-member notes — does it emit one note per participant, or one shared note you have to split by hand? 3. Golden Thread carry-over — does each member's note reference *their* treatment-plan goals, or just the group topic? 4. Billing metadata — does it capture start/stop times and participant count for CPT 90853 / 90849?
| Tool | Speaker separation | Per-member notes | Per-member Golden Thread | Group tier price |
|---|---|---|---|---|
| Twofold | Yes (labeled) | Yes | Yes | Included in standard plan ($19 first month) |
| Mentalyc | Yes (labeled) | Yes | Partial (per-member summary) | $89.99/mo Group/Couples |
| Upheal | Yes (labeled) | Yes | Partial | $99/mo Pro |
| Eleos | Yes | Yes | Yes (CCBHC-aligned) | Enterprise only |
| Blueprint | Limited | No (single shared note) | No | N/A |
Consent for groups Every participant must consent to recording, every session. The cleanest workflow:
1. Written consent at intake covering AI-assisted documentation. 2. Verbal reconfirmation at the start of each group, captured in the recording. 3. Any member declines → fall back to manual notes for that session. Do not record "just the others" — the audio still contains the declining member.
Where all group scribes still fall short - Cross-talk degrades speaker labels quickly. Expect to correct attribution on 5–15% of quoted content. - Silent members get thin notes. The scribe can only document what was said; you'll still need to add engagement/affect observations for quiet participants. - Process-level observations (group cohesion, sub-grouping, scapegoating dynamics) are not reliably captured by any tool — write these yourself.
Practical setup - Use a single omnidirectional mic in the center of the room for in-person groups, not laptop mics. - For telehealth groups on Zoom/Doxy, use the platform's cloud recording piped into the scribe rather than a laptop mic capturing speaker output. - Review per-member notes before they hit the EHR — group notes have the highest attribution-error rate in the category.
See also: SOAP vs DAP vs BIRP for group-note format choice, and hallucinations in therapy notes for the quote-fabrication pattern that's especially common in group transcripts.