Lyssn

The measurement is the product and it cannot exist without the model. Automatically estimating CBT fidelity from a session recording, identifying whether a clinician used specific evidence-based techniques competently, is not achievable by workflow software; the only alternative is a trained human rater listening to the session, which is precisely the bottleneck the company exists to remove. Founded by clinical researchers and data scientists rather than by operators adding AI to an existing service.

The design keeps the human decision explicitly in place while automating the measurement, and the company states it keeps the crucial human element front and centre. Fidelity scores feed supervision rather than replacing supervisors: the platform supports time-linked comments inside session playback and asynchronous review between therapist and supervisor, so the AI produces the evidence and the supervisor produces the judgement. That is a well-drawn boundary and it maps to the Ibex second-read principle, where the tool is positioned so it structurally cannot substitute for the professional.

The strongest transparency position in the behavioral health category and among the best in the index. The platform architecture is described component by component in peer reviewed literature, and the underlying methods are published in academic venues, including work on automated quality assessment of CBT sessions through highly contextualized language representations.

Most notably, the founders' equity stakes in the company are disclosed as competing interests within the papers themselves. A vendor whose science is published in venues that force conflict-of-interest disclosure is operating at a materially higher transparency standard than one publishing its own white papers, and the index should say so.

The system is described component by component in peer reviewed literature, which is a real and unusually durable form of disclosure: a published architecture cannot be quietly revised the way a product page can, and it lets a technical reviewer understand what the platform is composed of rather than accepting a category description. That answers more of this axis than most vendors manage. What it does not do is name anyone.

No foundation model provider, model class or version is identified, no hosting arrangement is published, and no sub processor list was located, so a buyer can describe the architecture and cannot name a single party that operates or supplies any part of it. The content raises the stakes of that gap above the category norm.

This platform holds full recordings of psychotherapy sessions, which is the most sensitive data type in this index, including sessions with people experiencing psychosis and substance use disorders, and no retention schedule or statement on whether recordings contribute to model training was located in either direction. A vendor willing to publish its architecture in the literature is well placed to publish who runs it. Ask for the sub processor list, the retention schedule, and an explicit training position.

Genuinely independent and academically rigorous, but still substantially forward-looking rather than concluded, which is why this is B rather than A. Project AFFECT is NIMH-funded, conducted with the Penn Collaborative for CBT and Implementation Science, and registered as NCT05340738 comparing AI-generated fidelity feedback against usual care. University of Washington work with King County demonstrated AI can accurately identify use of key CBT for psychosis skills.

A published body of academic papers covers fidelity measurement, supervision attitudes, opioid treatment programme counselling quality, and ClientBot. The honest caveat: much of the flagship work is protocol and feasibility stage, the psychosis skill detection study establishes that the AI can measure the skill rather than that measuring it improves patient outcomes, and the causal chain from better-measured fidelity to better patient outcomes remains the open question. Compare Eleos Health, which has a completed randomised trial with patient outcomes but a small single-site sample.

Described in the academic literature as a HIPAA-compliant cloud platform for secure recording, sharing and reviewing of therapy sessions, with the security posture documented in peer reviewed publication rather than only in marketing.

Graded B rather than A because no published retention schedule or statement on whether customer session recordings contribute to model training was located, and this platform holds the most sensitive data type in the index, full recordings of psychotherapy sessions including sessions with people experiencing psychosis and substance use disorders.

HIPAA compliance is documented in peer reviewed literature describing the platform, which is stronger corroboration than a marketing claim. Graded B rather than A because specific BAA terms were not located.

The earlier assessment located no attestation. That is overturned, and the position is stronger than a single report.

The vendor has completed a combined SOC 2 and health privacy attestation annually, announcing its fifth consecutive year, with the auditor named. It states that reports are available on request, describes encrypted transfer and storage, two factor authentication and continuous monitoring, and separately notes education records compliance for its public sector customers.

One distinction has to be got exactly right, and it runs in an unusual direction. The vendor consistently states Type 1 across its own announcements. A third party product directory describes it as holding SOC 2 Type II. The vendor is correct and the directory has inflated it. Credit belongs to the company for stating its report type accurately over several years, which several vendors in this index do not, and the aggregator's version should not be relied on.

What that means substantively: a Type 1 report assesses whether controls were suitably designed at a point in time. A Type II tests whether they operated effectively across a period. Five annual Type 1 reports are not equivalent to one Type II, because repeating a point in time assessment does not aggregate into period coverage. Annual repetition does give a recency benefit a single report would not, and it demonstrates sustained investment.

For a platform holding complete therapy session recordings, including crisis line calls in which suicidality is discussed, Type II is the report to ask for. Ask whether one is planned, and request the current report and its scope.

Not an FDA regulated product. The system measures clinician performance and supports training and supervision. It does not diagnose patients or direct treatment, so it sits outside Software as a Medical Device and the missing clearance is not the reason for this grade.

The C reflects the absence of a stated position on the oversight that does apply. The relevant surface is professional training and supervision standards, and for public sector deployments, procurement and quality assurance requirements. Nothing published sets out which standards the company believes govern its assessments or how its scoring relates to them.

That matters more here than for a documentation tool, because the output is an evaluation of a named clinician's practice. A buyer should establish what the scores are permitted to be used for, whether they can enter a personnel process, and what recourse a clinician has to contest one.

Ahead of the behavioral health field, though not complete. The published research programme includes work on automating the assessment of multicultural orientation through machine learning and natural language processing, meaning cultural responsiveness is treated as a measurable dimension of therapy quality rather than ignored.

A separate published study examined clinician knowledge and attitudes toward AI-based fidelity measurement in supervision, which takes the acceptability and fairness question seriously from the practitioner side. Graded B rather than A because no subgroup performance breakdown was located, and the stakes are high in a specific way worth naming: a fidelity score that systematically underrates clinicians with particular accents, dialects or therapeutic styles becomes an unfair performance evaluation attached to someone's professional standing.

The route into this band is publication, and this vendor publishes in a way that is structurally stronger than anything else in the index. Its methods appear in academic venues, including work on automated quality assessment of cognitive behavioural therapy sessions, so the approach is described where other researchers can examine and contradict it.

More significant than the publication itself is the venue's requirement: the founders' equity stakes in the company are disclosed as competing interests within the papers. That is a transparency standard imposed by a third party rather than chosen by the vendor, and it is the difference between a company that publishes and a company that submits to a process it does not control.

A reader of a vendor white paper has to assess the author's interest for themselves; a reader of these papers is told. Held below the top grade because none of it attaches to the product a buyer receives. No accuracy figure, error rate or evaluation methodology for the shipping platform was located as distinct from the published research, no warranty, indemnity or remediation commitment exists, and the platform holds full recordings of psychotherapy sessions including sessions with people experiencing psychosis and substance use disorders, where an error in a quality assessment attaches to a named clinician's professional record. Ask whether the deployed system is the one that was studied, and what a clinician can contest when an automated quality rating is wrong.

Browser-based platform where sessions are recorded directly or uploaded, operating largely as a standalone system for supervision and quality assurance rather than embedding in the clinical record. No named EHR integrations were located.

This is a coherent design choice given the buyer is a training or quality function rather than a documentation workflow, but it contrasts sharply with Eleos and Blueprint, both graded A for EHR-agnostic embedding, and it means the fidelity data lives apart from the clinical record.

Cloud platform accessed via web browser, supporting in-person and telehealth session recording as well as upload of sessions captured elsewhere, with user management organised around clinicians, supervisors and sessions. Flexible for the training and supervision use case. No data residency or hosting disclosure located.

No pricing published and no pricing basis disclosed. The buyer is an organisation, a training programme, health system or public agency, so pricing is presumably contracted, but nothing about the structure is public. Note the company's commercial position is also entangled with its research funding, since NIMH grants supported development of software the company intends to commercialise, a relationship the academic papers disclose openly but which buyers evaluating independence should understand.

Spans community mental health, opioid treatment programmes, psychosis and serious mental illness services, crisis and counselling services, and clinical training programmes, with modality-specific fidelity models for evidence-based practices including CBT and CBT for psychosis, plus motivational interviewing style counselling contexts. Public sector reach is real, evidenced by the King County healthcare system partnership.

Graded B rather than A because the product measures fidelity to specific named evidence-based practices, so coverage is bounded by which fidelity models have been built and validated, in the same way Unlearn's coverage is bounded by which Digital Twin Generators exist.