AI Medical Scribes Compared: What the Research Actually Shows

Ambient AI scribes record a patient visit, transcribe the conversation, and draft a clinical note without the clinician typing during the encounter. Adoption has moved fast: KLAS Research reports that 93% of health systems surveyed project moderate-to-deep ambient AI adoption, and more than 70% of large integrated delivery networks have already deployed or piloted a scribe tool. Nine vendors compete in this market with different deployment models, EHR depth, and target customers, from enterprise health systems to solo practitioners.

A comparison across these vendors runs into a consistent limit: most do not publish pricing. Commure, Nabla, Suki, Ambience Healthcare, and DeepScribe keep rates behind a sales quote. Where a dollar figure appears below, it is labeled as either the vendor's own published rate or a third-party estimate; the two carry different weight. Several of the review sites that publish competitor pricing estimates are themselves scribe vendors reviewing rivals, so their numbers are reported claims, not audited figures. Quoted complaints and praise are attributed to their original source; where a figure comes from a search snippet or a secondary aggregator rather than the vendor or a verified review platform, this post says so.

Commure Scribe (formerly Augmedix)

Commure Scribe is the ambient-documentation component of Commure's larger health-AI platform. Commure, with Athelas, acquired Augmedix, a 2013 ambient-documentation pioneer, for approximately $139 million ($2.35 per share), a deal that closed in late 2024, per Fierce Healthcare and Healthcare Dive. The product runs as a mobile app that records the visit; a third-party review notes there is no browser recording option, and direct EHR write-back is available only on the enterprise tier. The vendor claims more than 30 EHR integrations and a listing in Epic's Toolbox program.

Pricing per that same review: a Pro tier at $29/month billed annually or $79/month billed monthly, with enterprise pricing available only by quote; none of this is confirmed on Commure's own site. Commure announced a partnership with HCA Healthcare in October 2024 and a rollout across Tenet Healthcare's physician network in November 2024, per Healthcare Dive. Independent reviews of Commure Scribe are scarce. One third-party review site reports praise for support of 60+ languages, alongside complaints about Android app crashes and integration churn that the reviewer attributes to Commure's history of acquisitions.

Nabla

Nabla is an ambient scribe built by a Paris and New York company, available as a browser app, a mobile app, and an EHR-embedded option called Nabla Connect, built on SMART on FHIR. Nabla's site lists integrations with Epic, athenahealth, Oracle Health, NextGen, and Greenway, and claims more than 85,000 clinicians across 130+ organizations processing 20 million-plus encounters yearly. Named customers include Carle Health, Denver Health, and University of Toledo Health.

A free trial exists, but Nabla does not publish paid pricing. Third-party reviews and search aggregates cite a figure of roughly $119 per clinician per month, a reported estimate rather than a confirmed rate. Nabla raised a $24 million Series B in 2024, with Cathay Innovation named as an investor, and scored 90.7 (grade A) in KLAS Research's 2025 Ambient Speech ranking. Review site Twofold notes that Nabla's own claims, that 55% of users save one or more hours a day and that burnout drops 27%, are vendor-cited figures from supported deployments rather than independently audited results. Twofold's compilation of Reddit discussion places Nabla in a "mixed feedback, trade-offs" group rather than at either extreme.

Abridge

Abridge is an enterprise-only ambient scribe founded by cardiologist Shiv Rao, with the deepest Epic integration in the category, embedded across Epic's Haiku, Canto, and Hyperdrive interfaces. Funding has moved fast: $250 million in February 2025 at a $2.75 billion valuation, then a $300 million Series E in June 2025 at a $5.3 billion valuation, led by a16z with Khosla Ventures, per TechCrunch. Research firm Sacra estimates Abridge's annual recurring revenue at roughly $100 million as of May 2025, with $117 million in contracted ARR for the first quarter of 2025.

TechCrunch reports Abridge serves more than 150 large U.S. health systems; Contrary Research names Mayo Clinic, Duke, Johns Hopkins, and Kaiser among its customers. Abridge was rated Best in KLAS for Ambient AI in both 2025 (score 95.1) and 2026. There is no public pricing, no individual tier, and no self-serve trial; review site Vero estimates a cost range of $2,500 to more than $7,200 per clinician per year. Vero attributes a complaint sourced from Reddit and Capterra to users outside large systems: "Great product if you're at Mayo. Completely inaccessible if you're in private practice." The same source describes opaque sales pricing and a concern about 30-day audio retention. Abridge has also launched an AI medical coding product, per TechCrunch.

Nuance DAX Copilot / Microsoft Dragon Copilot

Nuance DAX Copilot is Microsoft's ambient documentation product. In March 2025, Microsoft merged DAX Copilot with Dragon Medical One into a single product called Dragon Copilot; Microsoft's health-solutions page confirms Dragon Copilot as the successor to Dragon Medical One. The product captures ambient audio with speaker differentiation across mobile and desktop, with EHR-embedded access; Microsoft's page references partnerships with Epic, athenahealth, and MEDITECH.

Microsoft does not publish pricing. Reseller DictationOne lists a rate of $600 per user per month plus a $650 setup fee for the first user; other third-party sources put the range at $369 to more than $830 per provider per month. Dragon Copilot scored 91.6 in KLAS's 2025 ranking, a grade of A-. Vero's roundup of source material, drawing on G2, Reddit, and forum posts, reports praise for maintaining eye contact with patients and for time savings per encounter, alongside a complaint pattern it describes as Reddit "consistently" framing DAX as too expensive outside large health systems. The same roundup notes verbose assessment-and-plan sections, ongoing edit burden, and quotes an observation that "DAX takes 6 months to deploy."

Suki

Suki positions itself as an AI assistant rather than a scribe alone: ambient note-taking plus dictation, voice commands, chart question-answering, and coding support across ICD-10, HCC, CPT, and E&M codes. Founded in 2017 and based in Redwood City, Suki runs on desktop and iOS/Android across more than 100 specialties. Suki's site claims real-time, bidirectional integrations, meaning the tool reads the chart and writes the note back, with Epic, Oracle Health, athenahealth, and MEDITECH, plus more than 400 health system and partner relationships.

Suki does not publish pricing on its own site. Third-party sources list a Suki Compose tier at $299 per user per month and a Suki Assistant tier at $399 per user per month, with enterprise pricing by quote, unverified against Suki directly. Suki raised a $70 million Series D in October 2024 and scored 92.9 (grade A) in KLAS's 2025 ranking, placing second in the Ambient Speech category; a 2024 KLAS spotlight reported that 95% of surveyed customers said they would buy again. A Reddit roundup compiled by Twofold credits Suki with accurate history-of-present-illness capture and well-organized assessment-and-plan sections.

Ambience Healthcare

Ambience Healthcare is an enterprise platform combining ambient documentation with point-of-care coding, a real-time ICD-10/CPT gap-check feature called AutoCDI, and inpatient clinical documentation improvement tools, built on OpenAI's models. The company joined Epic's Toolbox program in August 2025 and is embedded in Epic Hyperspace and Haiku; it also integrates with Oracle Cerner and athenahealth, per Fierce Healthcare and Ambience's own blog.

Ambience raised a $243 million Series C in 2025 at a $1.25 billion valuation, with Oak HC/FT, a16z, the OpenAI Startup Fund, Kleiner Perkins, and Optum Ventures participating. Pricing is sales-only, with no public numbers; third-party site UsagePricing estimates $2,800 to $5,000 per provider per year across modules, plus implementation fees. Reviewer Plaud cites similar figures alongside a reported six-month pilot at Cleveland Clinic that preceded a five-year contract, with 76% adoption within 15 weeks. Ambience scored 97.7 in KLAS's 2025 ranking, an unranked figure drawn from only 14 organizations. Its positioning emphasizes coding revenue over time savings alone; complaints from UsagePricing and Plaud center on enterprise-only design, implementation overhead, and integration depth concentrated on Epic.

Heidi Health

Heidi Health, founded in Melbourne in 2021, is a scribe with a genuine free tier, reportedly used in more than 50 countries with about 2 million weekly clinician sessions, per a Plaud review and BusinessWire. It runs as a web and mobile ambient scribe with note templates; EHR integration comes only at the Teams and Enterprise tiers.

Heidi's pricing page lists plan tiers without dollar amounts. Plaud reports a February 2026 restructure that raised the Clinician tier to $150 per user per month, about a 50% increase from a prior rate of roughly $99, and states a business associate agreement comes only with the custom-quoted Practice tier or above; treat this as reported, not vendor-confirmed. Heidi raised a $65 million Series B in October 2025, led by Point72, at a $465 million valuation, bringing total funding to roughly $96.6 million. In a KLAS Emerging Company Spotlight published in September 2025, 100% of surveyed customers said they would buy again, per BusinessWire. A Reddit roundup compiled by Twofold quotes a user, "grey-doc," describing Heidi as "almost perfect," with terse phrasing and a low hallucination rate; Plaud reports more recent feedback clusters around reliability, hallucinations, and support quality.

Freed

Freed is a self-serve scribe aimed at solo practitioners and small practices. Its pricing is published directly on its site: a Starter tier at $39/month covering 40 notes, a Core tier at $79/month with unlimited notes, and a Premier tier at $104/month billed annually or $119/month billed monthly, which adds a Chrome-extension push to the EHR, ICD-10/CPT coding, and patient letters. A Groups tier is custom-priced, a 7-day trial is offered, and the product is free for residents and students. Freed has no deep native EHR integration; the workflow relies on copy-paste or the extension push.

On the Apple App Store, Freed holds a 4.8 rating from 514 ratings. Visible reviews praise the tool for letting clinicians stay present with patients and for multilingual capture, and complain about bloated assessment-and-plan lists that require cleanup, battery drain, one report of data loss following an app update, and chatbot-first customer support. A Reddit roundup compiled by Twofold quotes the same user, "grey-doc," calling Freed "a real AI service. Quick and efficient," while noting some style drift between notes.

DeepScribe

DeepScribe is a specialty-focused ambient scribe founded in 2019 in San Francisco, having raised approximately $60 million per secondary sources. It is strongest in oncology: its site names Texas Oncology, Tennessee Oncology, Rocky Mountain Cancer Centers, and Ochsner as customers, and lists integrations with Epic, OncoEMR, and iKnowMed, plus an in-EHR mode called DeepScribe Embedded. It also offers SmartPrep pre-visit summaries, AI coding, and a Customization Studio.

DeepScribe does not publish pricing and offers no self-serve signup; third-party estimates put the cost at $350 to $500 per provider per month. It scored 97.6 in KLAS's 2025 ranking, though the figure is unranked and drawn from only seven organizations. A complaint compiled in Twofold's review of Reddit discussion describes DeepScribe as "horribly overpriced."

What the market-wide complaints show

Patterns that recur across nearly every vendor in this research point to where the category's real limits sit, regardless of which product a practice picks.

Hallucination and omission risk

A 2025 editorial in JMIR Medical Informatics, reviewing roughly 940 papers on ambient documentation tools, reports "potentially high rates of errors, including incorrect information, omissions, and hallucinations," and states that no standard evaluation method exists across the field. The editorial concludes that clinician oversight of every generated note is necessary, not optional.

Note bloat and edit burden

A physician-perspectives study published in PubMed found that clinician views on note accuracy and style were "largely negative, particularly regarding note length and editing requirements." The same study found that cognitive load, temporal demand, and work-life integration improved with ambient scribe use, meaning the tools reduced how draining the work felt even where the output itself needed more editing than clinicians wanted.

Cost at scale

Across vendors with any pricing data, enterprise-tier tools run roughly $250 to $830 per provider per month, whether published or estimated. Reddit discussion compiled by third-party sites calls out DAX Copilot and DeepScribe as overpriced specifically, and the enterprise-only sales models at Abridge and Ambience Healthcare exclude private practices from purchasing at all, regardless of budget.

EHR integration friction

Cheaper tiers, such as Freed's lower plans, rely on copy-paste or a browser extension rather than a true write-back into the EHR. Bidirectional write-back sits behind enterprise contracts across most vendors, and deployment can take months: Ambience's reported Cleveland Clinic pilot ran six months before a long-term contract followed, and one source quotes a claim that "DAX takes 6 months to deploy."

Privacy and consent

Reddit discussion summarized by third-party sites raises concerns about recording audio during sensitive encounters and about how long that audio is stored. One Capterra-sourced comment specifically flagged Abridge's 30-day audio retention period as a concern.

Environment sensitivity

Recurring complaint patterns across vendors include dependence on microphone and audio quality, processing delays during peak usage hours, and gaps in handling specialty-specific vocabulary.

Despite these complaints, adoption keeps climbing; KLAS's 93% adoption-projection figure suggests documented flaws have not slowed purchasing, which likely reflects how severe the documentation burden was before these tools existed rather than confidence that the flaws are resolved.

A framework for evaluating a scribe tool

The research above does not support ranking these nine products against each other; pricing gaps alone make that impossible, since some figures are vendor-published and others are third-party guesses. A practice evaluating this category can instead work through four questions before running a pilot.

• Specialty fit. Some products optimize for specific clinical contexts, such as DeepScribe's oncology focus and its OncoEMR and iKnowMed integrations. A tool tuned for a different specialty's vocabulary and documentation patterns may perform worse regardless of its general reputation.
• EHR integration depth needed. A practice that needs true bidirectional write-back into Epic, Oracle Health, or another EHR should expect that behind an enterprise tier and a sales conversation, not a self-serve signup. A practice comfortable with copy-paste or a browser-extension workflow has more self-serve options at lower prices.
• Budget tier. Published self-serve pricing exists at the low end, such as Freed's $39 to $119 monthly tiers, while enterprise contracts for products like Abridge and Ambience Healthcare are quote-only and, per third-party estimates, can reach the thousands of dollars per provider per year. Knowing the affordable tier narrows the vendor list before any product comparison starts.
• Pilot before scale. Given the documented hallucination and omission risk, and the note-length and editing complaints recurring across nearly every vendor, a short pilot with a small group of clinicians, checked against the actual editing time and error rate in that practice's own documentation style, is a more reliable input than any vendor's marketing claim or a competitor's third-party estimate.

None of the figures a vendor publishes about time saved or burnout reduction should be treated as verified unless the source is named and the study design is available for review. The KLAS scores here come from KLAS's own surveys of health system customers, making them a more consistent comparison point across vendors than pricing, which remains the least standardized part of this market.