Behavioral Health AI

Canary Speech sells the model rather than the application. Its vocal biomarker engine extracts roughly 2,590 acoustic and linguistic features from the human voice every 10 milliseconds and returns behavioral and cognitive indicators from as little as 20 to 45 seconds of natural speech, delivered through an API that other products embed. Canary Ambient is the API first real time offering, running on patient and clinician conversations, with a continuous monitoring variant added subsequently. Canary Cognitive addresses the cognitive side. Because the analysis reads how someone speaks rather than what they say, the company describes the approach as language agnostic and device agnostic, workable on anything with a microphone. The published evidence is unusual in this lane for what it admits. A 2026 paper in the Proceedings of Artificial Intelligence in Medicine reports unweighted average recall of 0.70 for depression and 0.68 for anxiety, with a combined behavioral health assessment reaching sensitivity of 0.76 and specificity of 0.65 on a remote test set, and sensitivity of 0.68 with specificity of 0.80 on an independent in clinic dataset collected by tablet. Those are moderate numbers and the company published them anyway, at an operating point, which almost no competitor here does. The same work introduces an uncertain classification that flags low confidence predictions near the decision boundary rather than forcing a call, which is a deliberate clinical safety choice rather than a performance figure. Markets span health systems, payers, pharmaceutical and clinical trial work, employers, telehealth and clinical call centres. In February 2026 the company partnered with Intermountain Ventures on an institutional review board approved study led by a neurologist at Intermountain Health, testing whether vocal features can identify multiple sclerosis. That same month it entered consumer health through JubileeTV, embedding passive speech analysis into video calls between older adults and their families, its first deployment outside clinical and research settings, producing indicators the company is careful to describe as nondiagnostic. Founded in 2017 in Provo, Utah by Henry O'Connell, who began his career at the National Institutes of Health in a neurological disease group before a long medical device career, and Jeff Adams, a speech recognition specialist. Roughly 22.35 million dollars raised in total, including a 13 million dollar Series A in June 2024 led by Cortes Capital and a further round in January 2025. Investors include Sorenson Communications, SMK Corporation, Plug and Play Japan and Hackensack Meridian Health, the last of which is a health system and therefore a party whose commercial and investment interests a buyer should hold separately when weighing any reference from it. One disclosure detail deserves attention before a security review. The company's security page lists an extensive set of assurance programs covering the SOC family, FedRAMP, FISMA and four ISO standards, and attributes them accurately to the cloud partners it builds on rather than claiming them. Canary's own stated position is HIPAA compliance, control audit against a recognised benchmark, vulnerability scanning and cleared API penetration testing. No attestation held by Canary itself was located. The wording is honest; the visual effect of the list is not, and a buyer skimming it may credit the company with certifications it does not hold.

Aiberry runs a bot administered interview and scores mental health from how a person answers it. A digital animated assistant called Botberry asks open questions, the person replies in their own words on camera, and an ensemble of machine learning models reads text, audio and video together to produce a depression risk score, symptom level insight across mood, concentration and energy, and a transcript of every response. The design intent is to replace the multiple choice self rating form, where a person picks the answer that best describes them, with something closer to a conversation. Two products are sold. Digital MindCare is anonymous and login free, aimed at schools, employers and community health organisations, and collects no account or identified health record. Smart MindCare, also presented as SmartAI MindCare, is the account based clinical product where personal and health information is collected and results land in a portal for a clinician. Four buyer settings are named: corrections, corporate wellness, behavioral health and recovery, and higher education. The evidence position rests on a single peer reviewed study, and it needs reading carefully. The paper appeared in the Journal of Affective Disorders in 2024, led from the University of Texas at Austin with the Georgetown University Medical Center and the University of Arizona, covering nearly 400 participants aged 18 to 74 who completed both a Botberry interview and a gold standard depression questionnaire. It reported performance comparable to the questionnaire and found no evidence of bias by gender, age or race. Press coverage presents this as an independent university validation. The paper's own declarations state that it was funded by Aiberry, that two authors are employed by Aiberry including the author credited with conceptualisation, methodology, formal analysis and supervision, and that a third is an Aiberry research coordinator. The disclosure is properly made in the journal. The framing built around it in press material is what diverges. A second currency problem sits beside it. The company research page still links a preprint rather than the published paper, so a reader following the vendor's own citation lands on the unreviewed version of work that has since cleared review. Based in Seattle and led by co chief executives Linda Chung and Johan Bjorklund with founding scientist Newton Howard. An 8 million dollar seed round led by Confluence Capital Group with Ascension AI participating was announced in March 2023. Named collaborators include Georgetown University, the University of Arizona and Advocate Aurora Health. Disclosure outside the clinical study is thin, and a buyer should expect to establish most of it privately. A dedicated pass located no security page, no external security attestation, no trust center, no named record system integration, no hosting or data residency statement, and no pricing of any kind. The company news feed has not been updated since 2023 and no funding round after the 2023 seed was located, so commercial trajectory cannot be established from public material in either direction.

Videra Health builds multimodal artificial intelligence for behavioral health, analysing what a patient says, how they say it, and how they look while saying it. The platform sends asynchronous video, voice or text check ins that patients complete on their own devices in roughly three minutes, and scores them against established clinical instruments. Three model families work together: language models reading transcribed speech for markers such as absolutist phrasing, word complexity and sentiment; acoustic models extracting speech rate, pause duration, pitch variation and articulation; and computer vision models reading facial movement, microexpressions and affective state. The company publishes a model table, which is rare in this index. Five named models each carry a clinical application, a reported area under the curve and a modality: PHQ-V for depression screening at 0.88 to 0.90, GAD-V for anxiety at 0.90, PCL-V for post traumatic stress at 0.85 to 0.90, EPDS-V for postpartum depression at 0.89, and a video model for tardive dyskinesia at 0.87 to 0.92. The video instruments were developed in house as video native versions of the standard written scales by the co founder and chief clinical officer. The product line spans the patient journey rather than a single moment. Front Door answers inbound patient inquiries around the clock and captures scheduling preferences, Intake Accelerator moves clinical history and paperwork to the patient's own time, Assess runs continuous check ins between and after visits, Sidekick Notes documents sessions ambiently, Group Notes documents group therapy while separating individual patient voices, and Sidekick Elevate returns communication pattern data to clinicians as coaching. A separate life sciences line applies the same capture infrastructure to clinical trials and post market studies. TDScreen, a tardive dyskinesia screening tool, is offered free to any provider or patient. Reported scale is more than 300 facilities and over one million patient interactions, with more than 100,000 notes generated and a claimed 94 percent completion rate on assessments. Named deployments include Discovery Behavioral Health, which launched a platform called Discovery365 in collaboration with Videra and Brigham and Women's Hospital, and work through the Medical Technology Enterprise Consortium aimed at preventing suicide related deaths among service members. Integrations are named with Kipu, Alleva, BestNotes, BlueStep, Zoom, Epic and Cerner. Academic collaborations are stated with the University of Utah and the University of Texas. Founded in 2019 in Orem, Utah. Roughly 8.6 million dollars is disclosed across a 3 million dollar seed in 2021 and a 5.6 million dollar second seed in May 2024, both led by Peterson Ventures, with Mercato Partners, Epic Ventures, Philo Ventures, Rose Park Advisors and OATV also named. Two cautions belong on the face of this record. The company describes itself as FDA registered and displays that phrase as a badge alongside its HIPAA and security marks. Establishment registration and device listing is a filing obligation, not a review of safety or effectiveness, and it is neither clearance nor approval. No clearance was located for any product. Separately, all five peer reviewed publications are authored by company staff, and the named scientific advisor is a co author on both tardive dyskinesia papers, so the evidence base is genuine, published and internally generated rather than independently replicated.

Behavioral health integration infrastructure sold to risk bearing organisations, health systems and health plans rather than to patients. Founded 2016 in Philadelphia by Chris Molaro, chief executive, and Adam Pardes, chief operating officer; one source gives 2017. Funding totals roughly 58 million dollars, including a 20 million dollar Series B led by Magellan Health and a 25 million dollar growth round led by SEMCAP Health, with a later private equity round. The premise is that behavioral health need surfaces in physical care settings that are not equipped to handle it. Three named products address that in sequence. BHIQ applies machine learning across claims, record system, admission discharge transfer and assessment data to surface undiagnosed and undertreated behavioral health conditions across a population. IntegrateBH, expanded in February 2026 with capabilities from recent acquisitions, operationalises those findings through risk identification, intelligent referral matching and care coordination, delivering patient data inside the provider's existing workflow. TxProgress measures treatment impact over time. One component distinguishes this record from every other behavioral health vendor in the index and should be read against them directly. NeuroFlow operates Response Services, a human staffed safety net for crisis moments, which the company describes as unique in its ability and willingness to address suicidality. The two patient facing conversational products built immediately before this one, Wysa and Limbic, publish no crisis escalation pathway at all. This company sells one. Named customers include Magellan Health, Trinity Health, Bozeman Health and the Department of Defense, with platform reach reported at roughly 15 million patients. Note that Magellan Health is both a customer and the investor that led the Series B, so that reference is not disinterested. The company was a finalist in a federal grand challenge to reduce veteran suicide, receiving 250,000 dollars and advancing to a second phase, which is competitive government evaluation rather than a marketing award. Security credentials are stated correctly and with their tiers: health specific certification at the implemented one year level, a controls report at the second type, and privacy compliance. A published example reports a large integrated health system stratifying more than 3,000 patients within four months. The gaps are measurement and commerce. No accuracy figure is published for the risk identification models, no peer reviewed evidence was located, and nothing about pricing is disclosed.

Conversational artificial intelligence delivering cognitive behavioural therapy, behavioural activation, dialectical behaviour techniques and mindfulness through text chat. Founded 2016 by Jo Aggarwal and Ramakant Vemati. Corporate geography is genuinely split and sources disagree: press releases carry Boston datelines, the operating entity is registered in the United Kingdom, and the company has substantial Indian operations, which is consistent with its stated compliance across American, British, European and Indian data regimes. Reach is the largest of any record in this index: between six and six and a half million users across more than 105 countries, with the company separately describing coverage of more than 11 million lives. The business now runs on two tracks. A freemium consumer application offers basic chat free with a premium tier adding unlimited use and human coaches. A provider facing line sells into health systems, with a gateway product reported at more than 185,000 patients in the national talking therapies programme and a hybrid product placing artificial intelligence between sessions and a human therapist in them. Institutional customers named by the company include the National Health Service, the Singapore Ministry of Health, Cincinnati Children's Hospital Medical Center, Aetna International, Accenture, Colgate-Palmolive and Swiss Re. The regulatory position requires precision, because third party sources get it wrong. Wysa holds Breakthrough Device Designation from the American regulator, granted May 2022 for an agent treating adults with chronic musculoskeletal pain and associated depression and anxiety. A designation is an agreement to expedite review; it is not clearance and it is not approval. At least one independent directory describes the company as holding an approved and cleared Class II device, which is incorrect. The company itself appended an editorial note to its own announcement stating that the designation applies to that specific indication and does not represent blanket approval across its products, which is unusually disciplined disclosure and the opposite of the credential inflation recorded elsewhere in this index. Evidence is genuine. An independent peer reviewed trial published in a medical informatics journal found the tool effective for chronic pain and associated depression and anxiety, more effective than standard orthopaedic care and comparable to in person psychological counselling. Separate peer reviewed work reports therapeutic alliance comparable to human therapist relationships. A competitive research grant of roughly 7.2 million dollars from a major medical research charity followed in February 2026. Total funding is around 35 million dollars. The trust posture is the strongest encountered in this index, and is set out on the security and stewardship axes. This record is best read against Limbic, built immediately before it: two conversational mental health products, one holding an actual device certification and the other a designation, with very different architectures of consent, anonymity and clinical oversight.

Clinical artificial intelligence for mental health, and the most heavily regulated and best evidenced record in this index. London based, founded 2017 by a team including Dr Ross Harper, chief executive; one source gives 2018. Limbic Access is an artificial intelligence triage and clinical assessment assistant that conducts a conversational psychological assessment supporting patient self referral, classifying the eight common mental health disorders treated by the national talking therapies programme at a stated 93 percent accuracy. In January 2023 it became the first and only artificial intelligence mental health chatbot in the world to obtain Class IIa medical device certification under the United Kingdom conformity regime, audited by a named certification body against clinical effectiveness, safety and risk management. Limbic Care is a clinical assistant for caseload management and continuous patient support, now including ambient scribe functionality that captures conversations, generates letters and standardises notes, carried under a separate Class I device classification. Deployment is at national scale within one health system. Tools have delivered assessments for more than 650,000 patients, across 66 percent of integrated care boards in England and roughly 45 percent of talking therapies services. Funding includes a 14 million dollar Series A led by Khosla Ventures with participation from Gaingels and the family office Illusian, alongside expansion into the United States. The evidence base is unlike anything else in this index. The company reports two publications in a leading peer reviewed medical journal: a 2024 study on improved accessibility for minority groups, and a randomised double blind clinical trial reporting that specialist trained clinical agents delivered cognitive behavioural therapy outperforming both human clinicians and general purpose large language models, built on a cognitive layer architecture the company calls the Limbic Layer that sits above foundation models. Recorded as the company states it. Neither paper was retrieved and read in this pass, and the trial result in particular is an extraordinary claim that a buyer should verify at source before relying on it. Trust disclosure is correspondingly detailed: information security and cyber certification, the national clinical risk management standard for health software manufacturers, a named data protection officer, individually enumerated patient data rights, a breach notification commitment, and a stated practice of stripping patient identifiable information from incoming referral records. One statement deserves particular attention and is treated carefully on the security axis: the company describes itself as compliant with the main information security standard while separately describing a different scheme as a held certification, and compliant is not certified. The gap across this record is commercial. No pricing, unit of charge or licensing basis was located anywhere.

Creyos, formerly Cambridge Brain Sciences and briefly Creyos Health, is a Toronto company selling digital cognitive and behavioural health assessment to clinicians. The platform administers twelve short computerised tasks, each mapped to a named cognitive domain covering memory, attention, reasoning, verbal ability and planning, and scores a patient against a normative database the company states holds more than 85,000 participants. Digitised versions of standard behavioural health questionnaires sit alongside the tasks, and a care planning capability was added in 2026. The tasks originate in three decades of research by Professor Adrian Owen, who built them at Cambridge, launched the web platform in 2009, and continued validating them at Western University's Brain and Mind Institute after winning a Canada Excellence Research Chair. The company states the tasks have been used in more than 400 peer reviewed studies. Applications span mild cognitive impairment and dementia screening, attention deficit assessment, concussion and traumatic brain injury, and research use. The company positions the product as a clinical decision support tool and describes the platform as registered with the FDA, which is a different and much weaker status than clearance.

Callyope is a Paris company building an audio and language foundation model trained specifically on psychiatric data rather than adapted from a general purpose speech model. Patients use a smartphone application for voice journals, short structured speech tasks, symptom self reports, or passive analysis of calls with designated caregivers, and the model combines that speech with smartphone sensor signals covering sleep and activity and with clinical information to produce continuous symptom scores between appointments. The clinician side adds case summarisation from uploaded records, consultation notes, referral and hospitalisation documents, and identification of gaps in medical coding. The founding team is a research team: Rachid Riad completed a doctorate at the Ecole Normale Superieure on automatic assessment of cognitive, linguistic and emotional disorders affecting speech in Huntington disease, and the company publishes at speech science venues, including two papers at Odyssey 2026, one on learning health related speech representations and one on approximate signal processing as a route toward homomorphic encryption for audio. A registered trial of voice based biomarkers for predicting schizophrenia relapse is running with an estimated 200 participants and completes in October 2026.

Genomind pairs a pharmacogenetic laboratory test with software that turns the result into prescribing guidance, with a centre of gravity in mental health. The assay is a clinician ordered buccal swab genotyped by quantitative PCR across 26 to 27 genes, split between 11 that affect how a drug is metabolised and 15 that affect how the body responds to it, returning results in three to five days and covering more than 700 medications. The Precision Health Platform then integrates that genotype with the patient's current medication list to flag gene to drug and drug to drug interactions and propose alternatives. The company was founded in King of Prussia, Pennsylvania by Dr Ronald Dozoretz, sells to individuals, prescribers, long term care facilities, accountable care organisations, employers and health plans, holds a contract supporting testing for United States active duty service members, and licenses its reporting suite to other laboratories so that its interpretation layer reaches patients whose samples it never handled. What distinguishes this record from everything else in the category is what the company did with its own evidence. It sponsored and co authored a multicentre randomised controlled trial of its own test in major depressive disorder, blinded to both participants and raters, and the trial found no significant improvement in the primary efficacy outcome. The company continues to cite that trial on its own science page.

Nudge AI is a behavioral health documentation platform for psychiatrists, psychologists, therapists, counsellors and psychiatric nurse practitioners, producing psychiatric evaluations, therapy progress notes and treatment plans, and reaching into level of care documentation that most competitors ignore, including partial hospitalisation daily progress notes for substance use disorder programmes written to Medicare requirements. Its transfer mechanism is more capable than the paste workflows common at this end of the market: a Chrome extension it calls EHR Copilot fills fields in the record with one click, including checklists, radio buttons and structured fields rather than dropping a block of text. It carries three named controls rare at this size, AES-256 encryption, audit logging and consent tracking, the last being a feature almost nothing else in this index ships. The reason to read its coding behaviour closely is that it does not only suggest codes: it identifies billing opportunities such as psychotherapy add ons and automatically generates the justifications for them, which is producing the evidence that supports a claim rather than recording what happened.

Nextvisit AI is a New York company founded by a practising psychiatrist and a software engineer, building documentation exclusively for behavioral health across psychiatry, psychology, therapy and community mental health, for psychiatrists, psychiatric mental health nurse practitioners, psychologists, therapists, licensed clinical social workers and physician assistants. Its architecture is multi agent, with models that reference prior records, track clinical change and assemble patient timelines rather than producing one note at a time. That longitudinal orientation is the point of the product: it visualises milestones, clinical progress and key life events in one continuous view and tracks symptoms, medications, treatments and risk factors across visits, which matters more in ongoing psychiatric care than in episodic medicine. Behavioral specificity shows in its tagging for Spravato, TMS and lithium monitoring, the last requiring serial level checks that a general purpose scribe would not know to follow. A mobile companion called ClinicalConcert handles documentation away from the desk.

TheraPulse is a small ambient documentation tool built for mental health clinicians, producing progress notes and summaries in around a minute across SOAP, DAP and BIRP formats. It accepts three input paths rather than one: live session recording, dictation after the session for clinicians who prefer not to record, or upload of an existing audio file. Its multi voice detection tags and separates speakers, which it positions specifically for couples, family and group therapy. Two things distinguish it in a crowded behavioral segment. It states plainly that recordings are used only for processing and are deleted immediately after transcription, which is the clearest deletion commitment located in this segment on the most sensitive audio in the index. And it publishes session volume based pricing rather than flat seats. It also publishes guidance on how to raise recording with clients, which is worth reading closely: the guidance affirms that recording is optional and consent based, and it also coaches clinicians on how to frame the request.

JotPsych is an ambient documentation product built exclusively for behavioral health and, unusually, covering its full scope rather than one corner of it: psychiatry, psychology, therapy and counselling, with 60 or more behavioral health specific note sections including psychiatric evaluations, medication management, therapy progress notes and intake assessments, plus templates for TMS, Spravato and polypharmacy. Coding suggestions cover ICD-10, CPT and DSM-5-TR, the last of which no general purpose scribe offers. It carries longitudinal context by referencing prior notes when drafting, accepts plain English editing instructions, and claims adoption across more than 10,000 behavioral health practices. It has since grown into a full practice platform bundling billing, credentialing, e-prescribing, scheduling and a website, positioned as one vendor replacing seven. That expansion is where a buyer should look hardest: JotPsych describes its notes as audit proof by design, carrying the detail, time and language payers want to see and documenting complexity to support the code, while its billing arm is paid a percentage of what it collects.

Supanote is an ambient documentation tool built exclusively for mental health clinicians, launched in 2024, and its depth inside that niche is the reason to look at it. It supports the full behavioral health note format set including SOAP, DAP, GIRP, BIRP, SIRP, PIR and PIE alongside custom templates, recognises more than 20 therapy modalities including CBT, DBT, EMDR and internal family systems, and handles couples and group sessions, which most ambient scribes cannot because they assume a single clinician talking to a single patient. Its Super Fill function autofills notes natively into SimplePractice and TherapyNotes, the two dominant practice systems in this market, and its personalisation engine claims to learn a therapist's voice and terminology from one session. Two things need care. The vendor states its models were trained on thousands of real therapy sessions and insurance approved notes without disclosing provenance or consent, and its own published pricing differs materially between its own pages.

AI-native EHR for mental health practice, and the vendor that repositioned this category's economics. Rather than selling documentation alongside a separate practice management system, Upheal bundles AI notes, scheduling, billing, treatment plans, compliance checking, client portal, forms, payments and a HIPAA-secure telehealth video room into one platform, and charges $1 per session capped at $69 per month per provider with no per-seat fees and no AI add-on. The company frames the model directly as charging for impact rather than for access, meaning a lighter caseload pays less, which inverts the flat-subscription convention. In 2026 it restructured from a Free, Starter and Premium ladder into this usage-based model, and features previously gated at Premium including Golden Thread treatment plans, the compliance checker and advanced analytics were folded into the paid plan. The Golden Thread capability is the clinically substantive differentiator: it maintains the documented continuity between a client's diagnosis, treatment plan goals and session-by-session progress notes, which is precisely what payers and auditors examine to confirm medical necessity, and the system links plan goals to progress notes automatically after each session. The compliance checker flags documentation gaps before submission to reduce clawback risk. Session capture is flexible across the integrated video platform, in-person sessions, text, dictation and manual upload, with over 170 preset note sections, session analytics measuring talking ratio, cadence and sentiment across the session timeline, and an AI assistant that can recall prior sessions and pull treatment plans. Notes export to external EHRs via browser extension and single-click transfer, and the company reports a bi-directional API and participation in EHR integration programmes. Compliance covers HIPAA with a published BAA, PHIPA for Canada, and SOC 2 Type II. Note on categorisation: sources disagree on whether Upheal is an EHR or a documentation tool, because it functions as both depending on whether a practice adopts it wholesale or exports notes elsewhere.

AI clinical documentation built exclusively for mental health professionals, covering therapists, counselors, social workers, psychologists and psychiatric mental health nurse practitioners. Notes are generated from recorded sessions, uploaded audio, dictation, transcripts or typed summaries, so recording is an option rather than a requirement, which matters for clinicians who will not record therapy. Output covers therapy-specific formats including SOAP, DAP, BIRP, PIRP, GIRP, PIE and SIRP plus a reported 100 or more custom templates, alongside treatment planning, session analytics, CPT code computation, and an Alliance Genie feature tracking therapeutic relationship dynamics across sessions. The distinguishing characteristic is compliance posture, which is unusually strong for a product sold largely to solo practitioners: HIPAA compliant, SOC 2 Type II audited, and PHIPA certified for Canadian practitioners, a combination no other indexed behavioral health vendor matches. The company states plainly that it does not train on clinician or client data and does not retain session recordings, which is the specific disclosure absent from Blueprint and Eleos Health, both indexed. Pricing is published and tiered from a Mini plan around $19.99 per month for 40 notes through Pro and Super tiers for higher volume solo clinicians, with a group practice plan around $49.99 to $59.99 per seat offering unlimited notes and admin visibility. Important sourcing caveat for readers: Mentalyc publishes extensive comparison content evaluating its own competitors, including several vendors indexed here, and that content is a useful factual source but is obviously self-interested. Where this record cites competitive positioning, it draws on independent third party reviews instead.

Digital musculoskeletal clinic pairing computer vision motion tracking with a clinical care team of physicians, physical therapists, nurses and health coaches. NYSE listed as HNGE. The technology core is genuinely substantial rather than a wrapper: the company acquired wrnch in 2021, the developer of a leading computer vision platform for measuring human motion using the same class of three-dimensional tracking employed in elite sport and motion capture for film, and stated at the time that the acquisition gave it the largest computer vision team in digital health. That became TrueMotion, which turns an ordinary smartphone camera into a 3D motion assessment tool, detecting 87 reference points across the body with no specialised equipment, and provides live audio and visual feedback on exercise form. In October 2025 the company launched Movement Analysis, which uses TrueMotion to capture joint angles, symmetry and endurance and combines those objective measures with questionnaires to produce HingeScore, plus Robin, an AI care assistant for members. It also published a set of AI care principles governing development across the platform. The strategic argument is worth understanding: almost all established MSK outcome measures rely on patient self-report, so an objective movement measurement captured at home is a genuine measurement contribution rather than a workflow convenience. The company also sells Enso, an FDA cleared wearable delivering electrical nerve stimulation for pain relief. Serves self-insured employers and health plans at large scale. Indexed with a clear scoping note: this is a care delivery organisation with a strong technology layer, not a software vendor, and the AI is graded on that basis. Sits alongside Sword Health, its closest competitor, also indexed.

AI that measures the quality of psychotherapy rather than documenting it, which makes it structurally different from every other behavioral health vendor in this index. The problem it addresses is real and well stated in the academic literature: health systems worldwide have spent enormous sums increasing access to evidence-based psychotherapies such as CBT, yet there has been no scalable way to evaluate whether those therapies are actually delivered competently, leaving quality and effectiveness largely unmeasured. Lyssn provides a HIPAA-compliant cloud platform for recording, sharing and reviewing therapy sessions with speech-to-text transcription, AI-generated fidelity and quality metrics, time-linked commenting inside audio and video playback for asynchronous supervision, and dashboards summarising performance across a caseload. Therapists can record in-person or telehealth sessions or upload sessions recorded elsewhere. The buyer is a training programme, supervisor, clinic or public health system rather than an individual clinician seeking time savings. Founded by PhD-level clinical researchers and data scientists out of the University of Washington, including David Atkins, Zac Imel, Michael Tanana and Shrikanth Narayanan. The evidence posture is unusual in this index: the underlying science is published in peer reviewed venues with the founders' equity stakes disclosed as competing interests in those papers, which is a higher standard of transparency than vendor-funded studies normally meet. Project AFFECT is an NIMH-funded research partnership with the Penn Collaborative for CBT and Implementation Science, registered as NCT05340738, evaluating AI-generated CBT fidelity feedback against usual care. Separate University of Washington work with King County's healthcare system evaluated detection of CBT for psychosis skills. Products extend to ClientBot, a patient-like conversational agent for training basic counselling skills.

AI documentation and clinical assistant built for mental health therapists. The Assistant listens to the session, transcribes it and drafts progress notes, treatment plans and session summaries, with note types covering individual, couples, family and group therapy as well as psychiatry workflows, and customisable templates for intake notes, progress notes, treatment plans and psychotherapy notes. It also surfaces pre-session and post-session insight, intervention suggestions, homework and assessment assignment, plus measurement based care tracking. The company reports more than 70,000 mental health professionals on the platform. Deployment is deliberately dual track: a browser extension and EHR integrations let it run alongside whatever system a practice already uses without migration, and the company separately offers its own AI-native EHR covering scheduling, billing, forms and insurance. In 2026 it repositioned around that free EHR as an on-ramp with the AI Assistant as the paid, usage-priced layer. Important contrast for readers: Blueprint and Eleos Health, both indexed, occupy the same clinical space from opposite ends. Blueprint is a bottom-up, self-serve product sold to individual therapists and small practices with published per-clinician pricing and enormous adoption; Eleos is a top-down enterprise platform sold to behavioral health organisations with peer reviewed outcome evidence and opaque pricing. Neither posture dominates. A buyer choosing between them is really choosing between commercial legibility and published clinical evidence, which is the more interesting comparison in this category.

Voice based AI purpose built for behavioral health, substance use disorder treatment, and field based care, marketed as CareOps Automation. Natural language processing captures the live session, or a short clinician summary, and produces structured progress notes while extracting clinical insight from the conversation itself. The product line extends past documentation: Eleos Compliance scans eligible progress notes to flag documentation errors before claim submission, Clinical Insights surfaces treatment effectiveness analytics, Eleos Outreach serves field based case managers and community outreach providers, and leadership dashboards report on staff activity, caseloads and population health. Multilanguage support recognises over 150 languages, with notes translated into clinically accurate English. Deployment is EHR agnostic via a lightweight browser embedding rather than a per-EHR build, which the company positions as its structural differentiator. More than 200 organisations reported using the platform. What distinguishes this record in the index is that the evidence is unusually deep for a documentation product and is published in peer reviewed venues rather than asserted: a randomised clinical trial in the Journal of Medical Internet Research, a machine learning study on therapy homework assignment, and a real-world accuracy study across more than 2.83 million micro-dialogues. The company also frames its own product as augmented rather than autonomous intelligence throughout. Sits adjacent to the index's ambient scribe cluster but is graded in behavioral health because the clinical insight and evidence-based-care layer, not the note, is the substantive claim.

Digital cognitive assessment and care planning platform. The flagship product, BrainCheck Assess, is a Class II FDA regulated device that delivers a standardized battery of digital cognitive tasks covering memory, executive function, attention, mental flexibility and processing speed, drawn from digital versions of established neuropsychological instruments including Stroop, Trail Making and Digit Symbol Substitution. It sits within a wider platform of more than 30 clinical evaluation tools, screeners and care planning instruments that clinicians can combine into custom protocols. Co-founded by neuroscientist David Eagleman. More than 400,000 cognitive assessments have been completed on the platform. The evidence base is comparatively strong and is published rather than asserted: a validation study of 99 participants reported at least 88 percent sensitivity and specificity separating normal cognition from dementia and at least 77 percent separating mild cognitive impairment from the other groups; a later study found a Pearson correlation of 0.77 against the Montreal Cognitive Assessment; and a 529 participant multi-site study conducted with BrainBox Solutions and registered as NCT04423198 assessed test retest reliability and concurrent validity against traditional paper and pencil tests. Remote self-administration has been reported as statistically indistinguishable from in-person coordinator led sessions. Important scoping note: the AI content of this product is limited. The battery is a digitized and normed set of conventional neuropsychological tests with an expanding normative database, not a machine learning diagnostic model, and the company states the platform is not intended as a stand-alone diagnostic tool and is designed to support clinician interpretation rather than replace it. Indexed for the assessment platform, and graded accordingly on AI Centrality.

Physician supervised virtual care company delivering asynchronous care through chat and voice messaging. Medical AI gathers context and provides initial evidence based guidance, and a board certified physician from an in house medical group licensed across all fifty states joins the conversation when diagnosis, prescription, or referral is required. Behind the interface sits a purpose built clinician cockpit and a self developed EMR designed for asynchronous care. Sold to payers, employers, and health systems as a front door to care.

Voice AI care management for health plans and at risk providers, originating in vocal biomarker research for detecting depression and anxiety and now productized as Sage, an AI care manager. The distinguishing technical claim is the Empathy Engine, patented vocal biomarker technology trained on millions of live clinical patient calls, which reads emotional state from voice in real time rather than only transcribing words, and adapts tone and approach accordingly. Sage conducts inbound and outbound member calls covering health risk assessments, post discharge follow up, risk profiling, care gap closure, program enrollment, medication reminders, and triage, and is positioned explicitly as added staffing capacity rather than a chatbot, reserving clinicians for complex high touch interventions. Integrations include Salesforce Health Cloud and, through an April 2026 partnership with HealthEdge, the Care Solutions suite drawing on GuidingCare member data and returning outcomes to the platform. Company reported figures, which it acknowledges are not independently audited, include roughly 60 percent reduction in administrative tasks, 6x faster program enrollment, and 4x return on investment. Named customers per the company website include UnitedHealth Group, Aetna, Duke Health, and Highmark. Founded 2017 by CEO Mainul Mondal; approximately $75 million raised including a $45 million round in June 2025 co-led by Salesforce, Khosla Ventures, and CVS Health Ventures.

AI native operating system for behavioral health clinics, combining a set of named AI workers with an integrated EHR, CRM, and revenue cycle management in one platform. Sarah, the admissions coordinator, answers inbound calls, texts, and faxes around the clock and verifies insurance in real time, with more than 100,000 patients booked reported. Emma, the clinical assistant, generates ICD-10 and CPT coded session notes within about 30 seconds and is reported to have supported more than 3 million patient sessions. Eric handles coding, claim scrubbing, submission, and denial management. Laura manages between visit engagement. The fully integrated platform launched April 2026, partners with athenahealth through the athenaOne Marketplace, and reports clinics booking 30 percent more patients with clinicians saving more than two hours daily, across hundreds of mid to large behavioral health clinics. Behavioral health operations are a distinct problem from general practice, and the specialization is the point. Backed by Y Combinator, Precursor Ventures, Pioneer Fund, Rebel Fund, and Expansion Venture Capital.

Global mental health platform for employers and health plans. The AI core is Precision Mental Healthcare: machine learning models analyze large data sets to match members to the right care modality and provider, with Compass, the company's mental health specific EHR, capturing every provider patient interaction and feeding measurement based care insights back to providers. Serves 800+ companies. Recovery outcomes were reported in a peer reviewed JAMA Network Open study (average time to remission 5.9 weeks, roughly 70 percent of participants reliably improved). The company published VERA-MH, a responsible AI framework for mental health. Buyer is the employer or health plan rather than the provider organization; indexed here on the standalone index where behavioral health is in scope.