Heidi vs Nabla vs Dragon Medical: physio note AI verdict

Nine minutes. That is how long clinical documentation takes per 15-minute physiotherapy appointment, according to multiple EHR studies published in 2025. For a 3-clinician practice running eight appointments each per day, that works out to roughly 3.6 hours of notes every single day: time that does not show up in the appointment schedule, does not generate additional revenue, and in most practices ends up being completed after hours.

Three tools dominate the conversation when practice owners in physiotherapy start looking at AI to recover that time: Heidi Health, Nabla, and Dragon Medical One. They are not equivalent products. They use different underlying technology, cost significantly different amounts, and suit different practice workflows. Choosing the wrong one means either paying for features you do not need or adopting a tool that does not fit how your clinicians actually work.

This analysis scores all three against the four CARE Framework dimensions that matter for a physiotherapy practice: Clinical Workflow, Administrative Burden, Readiness, and Economics. The verdict is not close.

The distinction that changes everything: dictation versus ambient AI

Before comparing specific tools, the most important distinction is architectural. Two of the three tools in this comparison (Heidi and Nabla) are ambient AI scribes. The clinician places a device on a desk or in a pocket, starts the appointment, and the AI listens to the patient-clinician conversation. At the end of the session, it generates a structured clinical note automatically, without the clinician stopping to type or speak.

Dragon Medical One works differently. It is a dictation tool. The clinician speaks the note aloud and Dragon transcribes what is said with high accuracy. The note quality and clinical reasoning depend entirely on what the clinician chooses to say. Dragon does not generate notes independently; it captures spoken words and structures them as directed.

For a physiotherapy context, this distinction matters more than in almost any other specialty. A physio performing a shoulder assessment, joint mobilisation, manual therapy, or dry needling cannot hold a microphone and dictate while treating the patient. The clinical contact is hands-on by nature. An ambient AI that listens and generates notes passively fits this workflow. A dictation tool that requires the clinician to narrate the note does not.

So what for you: if your clinicians are in physical contact with patients for the majority of the appointment, ambient AI is the correct technology category. Dictation may work as a post-appointment workflow, but it adds a separate step rather than removing one, and it depends on the clinician's ability to reconstruct the session in full after the fact.

Heidi Health: the free-to-start ambient option with physio-specific templates

Heidi Health is an Australian-founded ambient AI scribe serving clinicians across 116 countries and 300-plus specialties (Heidi Health, 2026). For physiotherapy practices specifically, Heidi offers pre-built templates: a Physio Comprehensive SOAP covering musculoskeletal, neurological, and rehabilitation assessments; a Physio SOAP Outpatient template for documenting progress; and a Physio SOAP In-Patient template for acute hospital settings. Clinicians can also build custom templates. The AI listens during the appointment and populates whichever template is active, generating the note without the clinician stopping treatment.

The pricing structure is the most significant factor for independent practices evaluating this tool. Heidi offers a permanently free tier (not a time-limited trial) as an ongoing product with access to AI note generation. The paid Clinician plan runs $150 per user per month (approximately £118 per month at June 2026 exchange rates), adding custom templates, extended AI actions, and priority support. For a 3-clinician physio practice that starts all three on the free tier and upgrades only when the value is clear, the initial monthly cost is zero. That is the entry-level economics of clinical note AI in 2026.

The free tier is sufficient for most independent physio practices trialling the tool. The upgrade question is really about how deeply the practice wants to customise templates and whether the additional AI actions (generating referral letters, producing patient summaries) justify the per-clinician spend.

In practical terms, the reported time savings are meaningful. Neil Aitken, owner of Kinetic Physio in Scotland, reported saving one to two hours per day on documentation after adopting Heidi's custom templates, according to the Heidi Health case study blog (June 2026). At a conservative estimate of one hour saved per clinician per day across three clinicians, that is 15 hours per week returned to patient care or recovered from the evening admin session.

On data compliance: Heidi Health is HIPAA-compliant for US practices and covers UK GDPR requirements. Any UK practice going beyond the free tier should request a signed Data Processing Agreement before processing patient data. This is standard practice for any AI tool in a clinical setting, not unique to Heidi.

So what for you: for a 3-clinician physio practice looking to trial clinical note AI with no upfront cost and minimal setup time (Heidi onboarding typically takes under an hour), Heidi is the logical starting point. The physio-specific SOAP templates are a practical advantage over general-purpose tools that require significant customisation before they are clinically useful.

Nabla: dual compliance and enterprise EHR integration

Nabla is a Paris-based ambient AI scribe with the strongest multi-framework compliance posture of the three tools compared here: HIPAA, GDPR, SOC 2 Type II, and ISO 27001 certified simultaneously (Bastion GPT comparison, June 2026). It supports 55-plus specialties and integrates natively with more than 20 EHR systems including Epic, Oracle Health, athenahealth, and NextGen. Nabla specifies that it does not store audio and does not train its models on patient data by default. Both policies directly address the most common clinical governance concerns about AI scribing tools.

On pricing, Nabla moved away from public per-month tiers in 2026 and now sells primarily via demo and contract, particularly for enterprise customers. The last publicly referenced Pro price was $119 per user per month for unlimited consultations and full EHR integration (Bastion GPT comparison, June 2026). A free tier exists at approximately 30 consultations per month, but this tier does not include a Business Associate Agreement (BAA), which means it is not appropriate for processing Protected Health Information under HIPAA. Any US practice evaluating Nabla should start from the paid tier to maintain compliance from day one.

Like Heidi, Nabla is an ambient AI scribe, so it fits the hands-on physio workflow correctly. The differentiation relative to Heidi is the dual HIPAA-GDPR compliance posture, the no-audio-storage policy, and the deeper EHR integration catalogue. For a practice operating in a regulatory environment where board-level sign-off is required on any new data processor, Nabla's compliance documentation is notably more complete.

The cost comparison for a 3-clinician practice: three Nabla licences at the last referenced Pro price run approximately $4,284 per year. Three Heidi paid licences at $150 per month run $5,400 per year. If the practice starts all three Heidi users on the free tier and only upgrades one, the Heidi cost is $1,800 per year. The gap is real and grows in favour of Heidi Health for practices that do not need the full Nabla compliance stack from day one.

So what for you: Nabla is the strongest choice when dual HIPAA-GDPR certification is a non-negotiable requirement, when the practice uses Epic or NextGen and wants tight EHR integration from the start, or when the data governance conversation requires more than a standard BAA. For practices that do not have those specific requirements, the pricing differential relative to Heidi's free tier is difficult to justify at the evaluation stage.

Dragon Medical One: where it fits and where it does not

Dragon Medical One, owned by Nuance (a Microsoft company), is the most established product in this comparison by longevity. It runs on UK-based Microsoft Azure servers, carries ISO 27001 certification, and meets NHS security standards (Crescendo Systems UK, 2026). It integrates with more than 200 EHR platforms and has a 30-year track record in clinical vocabulary recognition. NHS trusts in the UK and large health systems in the US have used Dragon for years. The product's credibility in healthcare documentation is not in question.

The pricing structure is the most complex of the three. Dragon Medical One costs $99 per user per month on a one-year term, with a one-time implementation and training fee of approximately $525 per new user (Voice Automated, 2026). For a 3-clinician practice signing annual licences, the year-one total is approximately $5,652 (three licences at $3,564 plus three setup fees at $1,575 plus any IT integration costs). Year two drops to $3,564, assuming no additional users or system changes.

The fundamental issue for physiotherapy practices is the product type. Dragon is a dictation tool. The clinician speaks the note and Dragon transcribes it accurately. This works well for desk-based specialties, for consultants who dictate letters after clinic, and for clinicians who have a well-trained dictation habit and want precise control over note content. It does not work naturally in a physiotherapy treatment session where the clinician's hands and attention are on the patient.

The workaround is a post-appointment dictation workflow: the clinician sees the patient, then spends three to five minutes after the session dictating the note to Dragon. This does reduce transcription time relative to typing from memory, but it does not remove the documentation step from the appointment itself. The ambient AI approach (Heidi, Nabla) eliminates the documentation step within the appointment because the note is generated during the session.

Dragon does hold a genuine advantage in one area: terminology accuracy and custom vocabulary. Clinicians who have trained Dragon on their preferred clinical language, specialist terminology, and specific documentation formats over months or years often report very high satisfaction with the output. The tool's accuracy on highly specialised clinical dictation is its clearest differentiator.

So what for you: Dragon Medical One fits physio practices where at least one clinician has an existing dictation workflow and finds it faster to dictate after appointments rather than rely on ambient AI. If you are migrating from Dragon Professional Individual to a clinical-grade product, Dragon Medical One is the logical upgrade path. If you are starting fresh with no existing dictation habit and no specific requirement for post-appointment dictation, the year-one cost and the workflow mismatch make the case difficult to build.

Scored against the CARE Framework

The CARE Framework scores tools across four dimensions: Clinical Workflow fit, Administrative Burden reduction, Readiness (setup and adoption), and Economics (year-one ROI).

Clinical Workflow (0 to 25 points): Heidi scores 22. The ambient AI model fits hands-on physio appointments directly, and the physio-specific SOAP templates mean the output is clinically structured without significant customisation. Nabla scores 21: same ambient model, slightly fewer physio-specific templates out of the box but strong EHR integration for practices on supported platforms. Dragon scores 14: dictation is a workflow mismatch for treatment sessions, requiring a separate post-appointment step.

Administrative Burden (0 to 25 points): Heidi scores 23. Notes generated during the appointment and completed immediately after means zero carry-forward documentation. Nabla scores 21: same model, additional EHR push reduces administrative steps for practices on supported systems. Dragon scores 15: reduces transcription burden compared to typing but adds a post-session dictation step that ambient AI avoids entirely.

Readiness (0 to 25 points): Heidi scores 24. Free tier removes financial risk from the trial. Setup takes under an hour. No EHR integration required to start. Nabla scores 18: demo-and-contract pricing adds friction at the evaluation stage; EHR integration setup takes time. Dragon scores 12: the $525 per-user implementation fee is not an obstacle, it is a prerequisite. Vocabulary training and workflow configuration for dictation take days to weeks before the tool reaches acceptable accuracy.

Economics (0 to 25 points): Heidi scores 25. Free to start; $150 per user per month at paid tier; year-one cost for a 3-clinician practice is $0 to $5,400 depending on how many users upgrade. Nabla scores 17: approximately $4,284 per year for three licences at the last referenced price; no free pathway into HIPAA-compliant use. Dragon scores 11: approximately $5,652 year-one for three clinicians (licences plus setup fees); highest year-one cost of the three tools.

Totals: Heidi 94/100. Nabla 77/100. Dragon 52/100.

The gap between Heidi and Dragon is not a close call for a physio practice. The gap between Heidi and Nabla is real, but it narrows substantially for practices with specific EHR integration requirements or a compliance environment that demands dual HIPAA-GDPR certification from day one.

The recommendation

For most 3-clinician independent physio practices that have not yet adopted an AI note tool: start Heidi Health on the free tier. Run a 4-week trial with at least two of your three clinicians using it consistently. Measure documentation hours in week one and compare them in week four. Set a threshold before you start: if the tool does not save at least 20 minutes per clinician per day, evaluate the reason before deciding whether to upgrade, adjust templates, or consider an alternative.

The evidence points clearly toward Heidi as the starting point. The free tier eliminates financial risk. The physio-specific templates eliminate the blank-template problem. The ambient model fits treatment sessions. The reported clinical outcomes (one to two hours of documentation time recovered per day in real-practice case studies) are the type of return that changes how a practice operates, not just how it documents.

Choose Nabla if your practice uses Epic or NextGen and wants day-one EHR integration, if a board or compliance officer requires simultaneous HIPAA and GDPR certification plus no-audio-storage policy in writing, or if you are a group practice where the $119 per user per month becomes comparable to Heidi's paid tier across a larger headcount.

Choose Dragon Medical One if you are migrating from Dragon Professional Individual to a clinical-grade product, if your clinicians have a strong existing dictation habit and find post-appointment dictation faster than ambient AI, or if you need a tool with deep NHS EHR integration and an established NHS procurement track record.

On the data compliance point: the Data (Use and Access) Act 2026 received Royal Assent on 19 June 2026 and brought updated provisions into force for data processors handling health information in the UK. All three tools have a credible compliance position. Getting a signed Data Processing Agreement in place before you go live is not optional regardless of which tool you choose.

If you want an independent analysis of which tool fits your specific workflows, your existing PMS, and your compliance requirements, the AI Opportunity and Growth Assessment produces a ranked recommendation with conservative financial projections for each option. It costs significantly less than a year of the wrong subscription. The first step is a free 20-minute discovery call to assess whether an audit is the right starting point for your practice.

Related: The 12-hour problem: what physio practices are paying for admin every week · From $0 to $150: the real cost of AI for an independent practice · Why 73% of AI tools go unused in healthcare practices