Upheal

AI-native by construction rather than by retrofit, and the pricing model makes the claim verifiable rather than rhetorical. The company charges per session with AI included and no add-on fee, explicitly contrasting itself with legacy platforms that bolted AI on as a paid upsell. Notes, treatment plans, the Golden Thread linkage, the compliance checker, session analytics and the client-aware assistant are all model-driven. A practice that stopped using the AI would be left with a thin practice management system.

Draft-and-approve, stated directly: the AI drafts while the clinician reviews, edits and approves. The compliance checker is an interesting inversion of the usual pattern, since it uses AI to audit the clinician's documentation for gaps before submission rather than only generating content, which places a second machine check between the draft and the payer.

Graded B rather than A because no edit-burden or acceptance-rate figure is published, and because independent user feedback reported elsewhere in this category describes accuracy problems including speaker identification errors and dialogue that did not occur, which if current would raise the real review burden above what the workflow implies.

Feature-level description is detailed, including over 170 preset note sections and analytics measuring talking ratio, cadence and sentiment along the session timeline, but no base model, training data provenance or accuracy methodology was located. Third party review notes that third-party proof remains thin relative to larger scribes and recommends requesting the security packet during evaluation, which is a fair characterisation of the overall disclosure level. Same C applied to Blueprint and Mentalyc in this category.

Nothing identifies any party in the chain: no model or model family, no foundation model provider, no hosting arrangement and no sub processor list was located in two passes, and no explicit retention or training commitment was found in either direction, which is the gap that matters in this segment and which one peer here closes in a sentence.

The instruments are real and are not a substitute: the vendor positions itself as a business associate with a published agreement and holds an independent security attestation, but a compliance posture describes obligations rather than naming who processes content. The data surface is larger than for an audio only scribe and should be sized accordingly.

The platform captures video as well as audio through its integrated telehealth room, so what moves through the unnamed chain includes the client's face and their home environment, which is identifiable in ways a transcript is not and cannot be redacted the way text can. The analytics layer adds derived content on top, since talking ratio, cadence and sentiment persist as data about the session separate from the note. Ask for a sub processor list, an explicit training and retention commitment, and retention stated separately for video, audio and derived analytics.

No peer reviewed study, controlled evaluation or published accuracy assessment located. Third party review explicitly notes limited third-party review volume compared with larger competitors. The session analytics produce quantitative measures such as talking ratio and sentiment, but no validation of those measures against clinical outcomes was located, which matters because analytics presented to supervisors carry an implicit claim to meaning. Consistent with Blueprint and Mentalyc; the whole scribe sub-lane is evidence-light relative to Eleos Health and Lyssn.

Positions itself as a HIPAA Business Associate with a published BAA, holds SOC 2 Type II and PHIPA compliance, and independent comparison identifies data privacy and client consent as the area of sharpest distinction between vendors in this category, indicating both companies compete on it actively.

Graded B rather than A because, unlike Mentalyc which states plainly that it does not train on client data and does not retain session recordings, no equivalent explicit retention or training-use commitment was located for Upheal, and the platform captures video as well as audio through its integrated telehealth room, which is a larger data surface.

Publishes a Business Associate Agreement and positions itself explicitly as a HIPAA Business Associate rather than merely claiming HIPAA compliance, and adds PHIPA compliance for Canadian practitioners. Publishing the BAA rather than producing it on request is the distinction: most vendors in this index state compliance without making the agreement available, and this is one of the few where a buyer can read the terms before committing.

SOC 2 Type II attestation specifically, corroborated by independent review rather than vendor claim alone, plus a security packet available on request during evaluation. Together with Mentalyc, this puts the two solo-practitioner-priced products in this category ahead of the enterprise platforms Eleos Health and Lyssn on published security attestation, which is a genuinely surprising inversion worth noting: the cheapest tools have the clearest certifications.

Converted from Not Rated. The prior scoping was correct, and this vendor engages the substitute regime more directly than its peers, which is why it grades above them.

No device pathway applies. Documentation, practice management and workflow tooling fall outside software as a medical device.

What governs is payer documentation standards, and the product targets them explicitly rather than incidentally. A compliance checker examines notes against payer requirements before they are finalised, and the golden thread capability addresses the specific documentation expectation that assessment, treatment plan, interventions and progress connect coherently across a course of care. That is the actual standard payers audit against in behavioural health, and building to it is a more substantive regulatory position than stating that no clearance is required.

That distinction is the grade. Most vendors in this category correctly observe that device regulation does not apply and stop there. This one identifies the regime that does apply and productises compliance with it.

Held at B rather than A for the same gap as its peers: no published position on 42 CFR Part 2, which imposes stricter confidentiality on substance use disorder records than the general privacy rule and reached its enforcement date in February 2026. A product built around documentation compliance is well placed to address it and does not.

Ask for the written Part 2 position, and what the compliance checker validates against.

No governance framework, subgroup analysis or bias evaluation located. The exposure here is broader than for pure note-drafting tools because the platform computes session analytics including sentiment and cadence and surfaces them for supervision and outcome tracking.

Sentiment analysis has documented performance variation across dialect, cultural expression and affect presentation, and a clinician or client whose speech patterns are read inaccurately by those models is mischaracterised in data a supervisor may act on. No published evaluation of that risk was located.

Two passes located no accuracy or error figure, no published limitations, no evaluation methodology and no warranty, indemnity or remediation commitment. One output category on this record deserves separate treatment from the note, because it is not documentation and nobody appears to be treating it as a measured claim. The platform produces session analytics including talking ratio, cadence and sentiment plotted along the session timeline.

Those are derived characterisations of both people in the room: a metric describing how much the therapist spoke is a performance measurement of a named clinician, and a sentiment curve is an inference about a client's emotional state produced by software.

Neither has a published accuracy figure, a validation basis or a statement of what it is calibrated against, and sentiment analysis is known to vary by speech style, affect presentation and cultural expression, which in this population is precisely the variation the metric is being read as clinically meaningful. Two questions follow that a therapist should ask before adopting.

What are these analytics used for beyond the clinician's own reflection, particularly whether they are visible to a supervisor, an employer or a payer. And what is the client told about being scored. Ask those, ask for validation of the sentiment output, and ask what the vendor commits to when a note is wrong.

Serves both sides of the EHR decision, which is why sources disagree about what this product even is. A practice can adopt Upheal as its full AI-native EHR covering scheduling, billing, forms, payments and client portal, or use it purely for documentation and push notes into an existing system via browser extension with single-click transfer.

The company additionally reports a secure bi-directional API and participation in EHR integration programmes, and states integration with EHR and telehealth solutions generally. Note the practical caveat from third party review: larger multi-site organisations demanding direct EHR write-back may find the current feature set restrictive, so the export path is smoother than true bidirectional sync for enterprise deployments.

Flexible capture across an integrated HIPAA-secure video room, in-person sessions, text, dictation and manual upload, with browser-based workflow. Strong fit for virtual-first practices given the built-in telehealth room. Graded B rather than A because third party review flags cross-device gaps including the absence of native Android parity, which constrains clinicians working outside the Chrome and desktop path, and no data residency terms were located.

The most legible pricing model in the behavioral health category and arguably the most买er-favourable structure in the index. $1 per session capped at $69 per month per provider, with everything included: AI notes, telehealth, scheduling, compliance checker, payments, forms, client portal, treatment plans, no per-seat fees and no AI add-on.

A free tier with unlimited feature-limited notes requires no credit card, and group practice pricing uses the same per-session math billed at practice level with a custom enterprise tier above. The usage-based structure means cost scales with actual caseload rather than seat count, so a part-time clinician pays materially less, and the cap protects high-volume users. The company also states its own model plainly, charging for impact rather than access. Note the honest competitive caveat from third party review: clinicians who consistently hit the $69 cap may find a flat-rate competitor cheaper, so the model favours variable and lighter caseloads.

Covers individual, couples and family therapy formats across in-person and virtual delivery, serving solo providers, small group practices and virtual-first clinics, with supervision and outcome tracking analytics extending it to small team use. Full practice management breadth beyond documentation.

Graded B rather than A because third party assessment identifies larger multi-site organisations as a poor fit given EHR write-back and device parity limits, and because coverage stays within outpatient psychotherapy without reaching the substance use disorder, case management or field-based settings Eleos Health serves.