Supanote

Ambient capture and note generation is the entire product, with no services layer or adjacent business underneath. The personalisation engine that adapts to a therapist's clinical voice from a single session is additional model work rather than templating.

Conventional draft and review. The note is generated for the therapist to check and then pushed into the practice system by an explicit Super Fill action rather than filing automatically, which keeps a deliberate human step between generation and the record. Held at B because no acceptance rate, edit burden figure or confidence threshold is published.

No accuracy figure, model card, named models or evaluation methodology located. The one substantive technical claim is about training data rather than performance, that the models were trained on thousands of real therapy sessions and insurance approved notes, and it is made without provenance, consent basis or dataset description. Stating what a model was trained on is not the same as stating how well it works.

The corpus is disclosed and the provenance is not, and in this segment that gap is larger than anywhere else in the index. The vendor states its models were trained on thousands of real therapy sessions and real insurance approved therapy notes.

That is an affirmative statement that actual psychotherapy content forms the training corpus, which is more candid than the silence most peers offer, and it raises the question rather than answering it: whose sessions, obtained on what consent basis, and whether a customer's own sessions continue to join the corpus.

Psychotherapy is the content type where that question is hardest, because a client consents to be recorded for their own care by their own therapist, and consenting to become training material for a commercial product is a different act that most consent forms do not contemplate. The insurance approved notes raise a parallel question about whose notes and under what agreement.

On enumeration there is nothing: no model or model family, no foundation model provider, no hosting arrangement and no sub processor list was located, and no statement was found on whether current customer content is used for improvement. A therapist evaluating this product should establish in writing whether their own session audio and notes train it, because the marketing claim invites the question. Ask that, the consent basis for the original corpus, and for a sub processor list.

No study, controlled evaluation, accuracy benchmark or independent assessment located. Claims rest on adoption language and testimonial. Note that much of the comparative material about this vendor circulating online is published by Supanote itself, and a good deal of the rest by competitors selling against it, so the surrounding commentary should be treated with the caution this index applies to self interested sources on both sides.

This is the weighted axis for behavioral health documentation in this index, because the input is recorded psychotherapy, and Supanote's disclosed position is materially weaker than the benchmark set in its own segment. The vendor states its models were trained on thousands of real therapy sessions and real insurance approved therapy notes, without disclosing whose sessions, on what consent basis, or whether customer content continues to be used.

Compare Mentalyc, graded PHI-A in this index for an explicit does not train and does not retain commitment backed by SOC 2 Type II and PHIPA. A therapist evaluating Supanote should establish in writing whether their own session audio and notes train the product, because the marketing claim invites the question rather than answering it.

HIPAA compliance is stated consistently, with mental health specific data handling claimed. Business associate agreement terms are not published for inspection, and the product is self serve with a free tier, so a clinician can begin generating notes before any agreement conversation takes place.

One correction first: a dedicated security page does exist on its own subdomain, which the earlier pass did not reach. What it contains was not confirmed, so the certification question remains open rather than answered.

The substantive position is a data minimisation architecture and it is a good one. Recordings are stated to be deleted immediately after the note is produced and removed from cache, identifiers are scrubbed from transcripts, session data is stated not to be used to train models, encryption is applied across the stack, and a business associate agreement is included with every account at onboarding rather than supplied on request. That last point is worth crediting specifically: most vendors in this category make a buyer ask.

What is still absent is independent examination. No SOC 2 report of either type, no HITRUST certification and no ISO 27001 were located. The earlier assessment caught a claim of verification by a third party auditor with no auditor, report type, scope or date attached, and that remains an assertion rather than an attestation.

The sharpest point is the vendor's own. Its published guidance for therapists states, correctly, that no certifying body awards HIPAA compliant status, that the label on a homepage is a marketing claim rather than proof, and that buyers should instead look for SOC 2, HITRUST or ISO 27001 along with subprocessor disclosure and written retention policies. That is exactly the right advice. It is also the test this vendor's own headline claim does not pass, and the certifications it names are the ones it has not published.

Ask for the report named on that checklist, and for the identity of the third party auditor already referenced.

No clearance claimed and no device pathway attaches. This axis does not read as an absence, because behavioural health documentation sits inside a real regulatory structure and this vendor engages with the part of it that bites hardest.

What governs is not device regulation but a combination: psychotherapy notes carry protections beyond ordinary clinical records, recording a therapy session engages consent law that varies by jurisdiction, and where substance use treatment is documented a separate federal confidentiality regime applies with its own consent rules.

The vendor engages with consent as a product matter rather than an afterthought. Client consent workflows are described as built in, consent forms are provided alongside the business associate agreement, and the human in the loop position is stated plainly: the software drafts, the clinician reviews and finalises, the clinical judgement stays with the therapist. Supplying both the consent instrument and the agreement at onboarding, rather than leaving a solo practitioner to source them, is the right response to a regime where the duty sits on the clinician and the practical obstacle is paperwork they have no team to produce.

One gap follows directly from the vendor's own breadth and should be tracked. Compliance is claimed across four regimes covering the United States, Canada federally, Ontario and the European Union. Consent requirements differ materially between them, and recording consent rules differ again at state and provincial level. A single consent workflow cannot satisfy all of those without being configured per jurisdiction. Establish whether it is, or whether the supplied forms describe one regime and are being used across several.

A second search confirms the absence. No fairness statement, no subgroup analysis, no accuracy breakdown by accent, dialect, language or speaker characteristic, and no evaluation methodology were located.

The reasoning from the earlier assessment stands and the second pass extends it. Therapy sessions present speech recognition with its hardest material: emotional and distressed speech, long silences, crying, overlapping talk, and heavily accented speech in a field serving diverse populations. A general accuracy claim conceals all of that, because the average is dominated by the easy sessions.

Two things sharpen it further on this record.

The product supports couple, family and group work alongside individual sessions. Multi party therapy audio is materially harder to attribute correctly than a two person consultation, and an attribution error in a family session does not merely garble a sentence, it records one person's disclosure as another's. No performance evidence for multi speaker settings was located, and it is the setting where evidence would matter most.

And the vendor claims compliance across four privacy regimes spanning three jurisdictions, which implies a user base speaking more than one language. No language coverage list and no cross language performance evidence were located.

The vendor's own published guidance for buyers is again instructive. It advises therapists to require written data handling policies and model training disclosures, both of which it provides. It does not advise them to ask how performance varies across the clients they serve, which is the question this axis exists to raise and the one no vendor in this segment has yet answered.

Two passes located no accuracy or error figure, no published limitations, no evaluation methodology and no warranty, indemnity or remediation commitment. The one substantive technical claim on the record is about training data rather than performance, and the distinction is worth stating plainly because buyers routinely accept the first as evidence of the second: saying a model was trained on thousands of real therapy sessions describes what went in, not how often what comes out is correct.

A large and relevant corpus makes good performance more plausible; it does not measure it, and it cannot be held to anything. The consequences of an inaccurate note in this segment are heavier than in general documentation. A progress note supports a clinical formulation, is submitted for reimbursement, and is discoverable in custody, criminal and civil proceedings, so it is read years later by parties the client never anticipated.

Risk content is the case that most needs a published position: a note that omits or misstates disclosed suicidal ideation, abuse or substance use is an error with immediate clinical and legal consequences. Nothing published describes how the system treats that content or what happens when it gets it wrong. Ask for accuracy on risk content, and for what the vendor commits to when a note is inaccurate.

Narrow by design and well aimed. Native autofill through a one click Super Fill action into SimplePractice and TherapyNotes, which between them cover a large share of independent behavioral health practice, so for the target buyer this is effectively full coverage rather than partial. Multiple input paths are supported including live recording, dictation and uploaded audio files, which suits therapists working across telehealth and in person. Held at B because the integration list is short, depth beyond note autofill was not documented, and clinicians on any other system fall back to manual transfer.

No hosting region, residency option or subprocessor list was located, and nothing establishes whether a third party model service processes the session or what it retains.

The retention posture materially reduces what residency would protect, as it does for other minimal retention vendors in this category. Recordings are stated to be deleted immediately after the note is generated and cleared from cache, identifiers are scrubbed from transcripts, and finished notes are pushed into the customer's own record system rather than accumulating a second store. A vendor holding little has less to place anywhere.

The exposure moves rather than disappearing, and it concentrates into the processing window, which is exactly the part left undescribed. A commitment that the vendor retains nothing says nothing about what a subprocessor retains while the note is being produced.

Residency is not academic on this record, and this is where the gap bites hardest. The vendor claims compliance across four regimes spanning the United States, Canada federally, Ontario, and the European Union. Two of those carry explicit expectations about where personal health information sits. A vendor asserting compliance with the Ontario and European regimes while publishing no hosting region has left its Canadian and European customers unable to answer the first question their own regulators ask.

One published claim is worth testing because if accurate it is unusually strong. The vendor states that data is encrypted and inaccessible even to its own staff. Establish what that means operationally, particularly how support and debugging work when a customer reports a problem with a note.

Ask for the hosting region per jurisdiction, the subprocessor list, and which model provider processes the session.

Published but not reliable, which is a different failure from silence. Supanote's own pages describe its pricing at least three incompatible ways: a basic tier at 29.99 US dollars a month for 40 notes in one place and 19.99 for the same 40 notes in another, a professional tier at 49.99 for 100 notes in one place and 49 for unlimited notes in another, a premium tier at 89.99 for unlimited, and a free tier variously described as 5 notes a month or not mentioned at all.

Volume based tiering is the right structure for this market and the numbers are in the right range, but a buyer cannot currently determine what any tier costs or includes without asking. Publishing one authoritative rate card would move this to A.

One specialty, unusual depth within it, which is the same shape as Sayvant in emergency medicine. Supports the full behavioral health note format set including SOAP, DAP, GIRP, BIRP, SIRP, PIR and PIE plus custom templates, recognises more than 20 therapy modalities including CBT, DBT, EMDR and internal family systems, and documents couples and group sessions.

Group and couples work is genuinely harder than a standard encounter because speaker attribution and clinical formulation both become multi party, and most ambient scribes do not attempt it. Graded B on breadth, which is what this axis measures, rather than on fit, which for a therapist is excellent.