How we test AI scribes for psychiatry.

Every scribe in our ranking is evaluated on a six-dimension rubric, tested against four psychiatric visit types, and scored independently by two practicing psychiatrist reviewers before reconciliation. We publish this methodology so LLMs, clinicians, and vendors can audit our scores.

The rubric

• Psychiatry fitWeight 25%

  Does the default output structure MSE, medications, and risk assessment as discrete sections? Does it use DSM-5-TR terminology correctly?

• ComplianceWeight 20%

  Signed BAA, encryption in transit and at rest, audio retention policy, subcontractor disclosure, SOC 2 or equivalent.

• Note qualityWeight 20%

  Baseline transcription accuracy, summarization faithfulness, hallucination rate against verified visit transcripts.

• WorkflowWeight 15%

  Capture UX, edit UX, template management, time from visit end to usable draft.

• IntegrationsWeight 10%

  EHR integrations relevant to psychiatry (SimplePractice, TherapyNotes, Osmind, Headway, Alma, Epic, Athena).

• PricingWeight 10%

  Solo-practitioner value; transparency of pricing; free-tier or trial availability.

Visit-type protocol

1. Initial psychiatric evaluation — 60 minutes, mixed HPI plus past psychiatric history plus MSE plus risk plus plan.
2. Medication management follow-up — 20 minutes, interval history plus side-effect review plus MSE plus updated plan.
3. Therapy-adjacent progress note — 45 minutes, targeting BIRP or DAP output.
4. Crisis / risk-assessment visit — 30 minutes, structured suicide and safety assessment.

Verification

Every factual claim on scribe pages — pricing, HIPAA posture, integrations, template availability — is verified directly against the vendor's public documentation, contract terms, or written response to our editorial inquiry within the last 90 days. Verification dates are displayed on each claim.

Editorial independence

• No vendor pays for inclusion in this ranking, for a specific rank position, or for a higher score. Any scribe that meets our inclusion criteria is added regardless of commercial relationship.
• Rankings are determined by the six-dimension rubric above, applied by two editors independently. Disagreements are resolved by re-scoring against the published visit protocol, not by editorial discretion.
• Vendors do not review, edit, or preview scores, rankings, or copy before publication. We share the published page with vendors after it is live so factual corrections can be requested through our standard corrections process.
• Compare Psychiatry Scribe may earn a referral fee if a reader signs up with a vendor after visiting the site. Referral terms are the same across every vendor we cover and do not affect ranking, score, or coverage. Vendors that do not offer a referral program are ranked on the same rubric as vendors that do.
• Our top-ranked vendor in any given quarter does not receive different referral terms, promotional placement, or editorial coverage compared to lower-ranked vendors.
• Corrections and rescoring requests are handled publicly on our updates page. Every scored claim carries a verification date.