Choosing an AI scribe for physical therapy used to mean picking a transcription convenience. In 2026, it means picking infrastructure. Therapists spend roughly 8–15 minutes documenting every encounter, and across a 15–25 patient day, that adds up to 2–6 hours of charting layered on top of direct patient care. For a solo clinician, that's a bad evening. For a multi-site group running dozens of clinics and hundreds of providers, it's a structural drag on capacity, revenue, and staff retention — multiplied across every location, every payer, every week.

That's the audience this guide is written for: practice owners, clinical directors, RCM leaders, and IT decision-makers evaluating the best AI scribe for physical therapy across a network of clinics — not a single-provider tool comparison. In 2026, the conversation has moved past "does it save time" and into a much more consequential question: does the technology sit natively inside your operating system, or is it another point solution your staff has to babysit?

This guide covers what a PT-specific AI scribe actually needs to do, the compliance mechanics that separate a real solution from a transcription tool with a PT label on it, how the 2026 market is organized — including mobile and phone-based ambient capture — and what a network-level evaluation should weigh before signing a contract.

Quick Answer: Top AI Scribes for Physical Therapy in 2026

The rest of this guide explains why the market splits this way, and what actually separates the top of that list.

What Is an AI Scribe for Physical Therapy?

An AI scribe for physical therapy is a clinical documentation system that listens to (or is dictated) a patient encounter and converts it into structured, billable clinical notes — but unlike a general medical scribe, it has to understand rehab-specific data: range of motion in degrees, manual muscle testing (MMT) grades, special test results, gait and functional mobility status, and treatment time tied to specific CPT codes.

A general-purpose AI scribe built for a physician's SOAP note (chief complaint, history of present illness, exam, assessment and plan) will often miss or flatten this data. Physical therapy notes are measurement-driven and progress-oriented in a way that a generalist model, trained mostly on physician encounters, simply wasn't built to capture correctly the first time.

The American Physical Therapy Association formalized this distinction in 2025, publishing a practice advisory on AI-enabled ambient scribe technology that outlines documentation responsibilities, regulatory considerations, and the evidence behind the technology's benefits and limitations — a signal that ambient scribing in PT is now mainstream enough to need its own professional guidance, not a borrowed medical framework.

The 2026 Compliance Landscape: What a PT-Specific Scribe Actually Has to Handle

This is the part most generic "AI scribe" content skips — and it's the part that determines whether a tool is safe to deploy across a multi-site network or just a transcription convenience for a single provider.

The 8-minute rule. Timed CPT codes — 97110 (therapeutic exercise), 97140 (manual therapy), 97530 (therapeutic activities), 97542 (gait training) — are billed in 15-minute units, but the conversion from minutes to units follows a specific threshold table. Incorrect unit calculation is one of the most common PT billing errors and a frequent audit trigger, and it only gets riskier at scale, when dozens of providers are each making the same small judgment call dozens of times a day.

KX modifier and the therapy threshold. When a patient's therapy spending crosses the Medicare cap threshold — $2,330 for PT/SLP combined in 2026 — continued treatment requires the KX modifier, which certifies medical necessity and functional progress. Documentation has to clearly support that claim, because insufficient KX justification is a known audit target that can trigger payment recoupment.

Plan of Care (POC) certification. A signed plan from the referring physician — diagnosis, long-term and short-term goals, interventions, frequency and duration — is required, with re-certification every 90 days.

Medicare progress notes. Required every 10 visits or 30 calendar days, whichever comes first, documenting progress toward each goal and justifying continued skilled care rather than maintenance therapy.

The table below lays out the 8-minute rule thresholds directly, since it's the single most common compliance question multi-site billing teams ask.

The 8-Minute Rule — Timed CPT Code Billing Thresholds (2026)

When a therapist provides multiple timed services in one visit, the "remainder rule" applies: total minutes across all timed codes are summed to determine total billable units, and those units are then allocated to whichever codes consumed the most time. A scribe that doesn't track this automatically leaves it to individual providers to calculate correctly — every single visit, across every single location.

Why "Bolted-On" AI and "Native" AI Are Not the Same Category

This is the distinction that matters most for a network evaluating platforms in 2026, and it's where the market has genuinely split into two categories.

Bolted-on AI scribes are add-ons layered onto an existing EMR — either a third-party plug-in or a feature a legacy vendor has recently introduced to stay competitive. The transcription happens, but the output often lands as a block of text a provider still has to copy, reformat, or manually reconcile with structured EMR fields, billing rules, and prior authorization requirements. The AI and the operating system are two separate products wearing one interface.

Native AI scribes are built inside the EMR's actual data model from the start. The documentation output isn't a paragraph to copy — it writes directly into structured fields (MMT grids, goal trackers, dropdowns, nested conditions), and because it shares the same data layer as scheduling, billing, and payer rules, the note doesn't just get written — it can carry the visit forward into a clean claim without a second manual step.

For a solo provider, the difference might be a matter of convenience. For an enterprise group processing thousands of notes and claims a month across multiple locations, it's the difference between an efficiency feature and an operating system.

This is where SPRY sits deliberately in the second category. SPRY's AI Scribe Agent (native, launched April 2026) runs on a four-layer architecture — a perception layer that reads speech, documents, and forms; a decision layer that predicts coding, denial risk, and eligibility issues; an automation layer that executes the resulting workflow; and a learning loop that improves from real clinic outcomes over time. It briefs a provider on a patient's history before the visit, writes directly into the EMR's actual form fields rather than a copyable paragraph, accepts corrections in plain conversational language, flags clinical inconsistencies before a note is signed, and — because the note lives inside the same system as billing — initiates the revenue cycle the moment it's complete. Across real deployments, that has translated to initial evaluations completed in roughly 5 minutes and follow-up notes in about 2, versus documentation workflows that can drag on for hours in bolted-on systems.

Using an AI Scribe for Physical Therapy on a Phone

Most AI scribe for physical therapy tools in 2026 run through a phone or tablet placed in the treatment area, since a therapist's hands are occupied during manual therapy, gait training, and exercise instruction — typing during the encounter isn't an option. The phone or tablet's microphone passively captures the session, and the audio is processed into a structured note without the therapist touching the device again until it's time to review.

For a solo provider, this usually means a personal or clinic-owned phone running a scribe app. For a multi-site network, phone-based capture raises questions a single-device setup doesn't: is the app HIPAA-compliant and BAA-covered on every device, is audio processed and deleted according to a consistent retention policy across every clinic, and does the note land back in a personal device's app or directly in the shared EMR? A native, EMR-embedded scribe answers all three consistently across every location, because the mobile capture and the EMR are the same system. A bring-your-own-app scribe answers them differently clinic by clinic — which is a governance risk a network compliance officer, not just an individual provider, needs to own.

Comparing the 2026 AI Scribe Landscape

The table below reflects how the market breaks down across the criteria that matter to a multi-site evaluation — not just documentation speed, but whether the platform can actually operate as connected tissue between clinical work and the revenue cycle.

AI Scribe Landscape for Physical Therapy — Enterprise Evaluation Criteria (2026)

A few things stand out reading this landscape as a network operator rather than a solo buyer. Below is a closer look at where each of these tools actually fits.

DeepCura is the broadest platform on this list — ambient scribing bundled with an AI receptionist, billing automation, and EHR integration across 50+ medical specialties, priced at roughly $129/month per provider. Its physical therapy templates are capable, but PT is one of dozens of specialties it serves, not the platform's core design center.

ScribePT is a PT/OT/SLP-only documentation tool at roughly $99/month, built around native integration with WebPT and HENO. It's a strong fit for a single clinic already on one of those EHRs, but it stops at documentation — no billing, scheduling, or RCM automation, which means the note still has to hand off to a separate system.

WebPT + Comprehend Health pairs the largest install base in the category (90,000+ therapists) with a third-party AI scribe added via an August 2025 partnership. The scale is real, and the published case studies are credible, but the AI layer was integrated onto an existing EMR rather than built into it from the start — the documentation and billing workflows remain separate products.

Prompt Health (Sidekick) owns its AI scribe technology outright, acquired in 2025, and keeps documentation and billing inside one platform — a genuinely native model at the mid-market tier. It's a serious platform to evaluate, though its published outcomes are largely vendor-reported rather than third-party verified.

Ensora Health (formerly Fusion/TheraNest) carries a large historical install base but is still in the early stages of layering AI onto what is fundamentally a legacy EMR, with billing as a separate long-standing product line.

OneChart, Twofold, Freed AI, and Heidi Health sit at the budget end of the market ($39–$99/month), built primarily for individual providers or small independent clinics. They're reasonable entry points for a solo practitioner testing AI documentation for the first time, but none offer meaningful billing, eligibility, or multi-location governance features.

Nuance DAX (Microsoft) is the enterprise standard for hospital systems already running Epic, at $369+/month per provider — the right choice for large institutional PT departments, but priced and built for hospital procurement, not independent outpatient networks.

Most of the category — DeepCura, ScribePT, Twofold, OneChart, Heidi Health, Freed AI — is built and priced around individual providers or small independent clinics. That's not a criticism of those tools; it's simply a different buyer. A documentation-only scribe that doesn't touch billing, eligibility, or prior authorization is solving one piece of a problem that, at enterprise scale, has five or six connected pieces. The platforms built to compete for network-level deployments — SPRY, WebPT's Comprehend partnership, Prompt Health, and legacy systems like Ensora and Raintree — are the ones actually built around that broader operational question.

What a Multi-Site Rehab Network Should Actually Evaluate

If you're responsible for documentation and billing outcomes across more than one clinic, the evaluation criteria look different from a single-provider buying decision. A few questions worth asking of any vendor:

Where the Category Is Headed: Agentic Documentation

The next shift already underway in 2026 is the move from "ambient scribe" to agentic documentation — systems that don't just transcribe a conversation but carry memory and context across visits, reason about what's clinically relevant, and take action rather than simply producing text. An agentic system can brief a provider on a patient's trajectory before they walk into the room, apply what it learned about a provider's documentation style from session one to session two hundred, and flag a clinical inconsistency — a body-part mismatch, a missing functional limitation — before a note is ever signed, rather than after a payer catches it six weeks later.

This distinction matters because it's where the compliance and revenue-cycle stakes actually live. A transcription tool reduces typing. An agentic system reduces the downstream risk of an incomplete or inconsistent note ever reaching a payer in the first place — which is a fundamentally different value proposition for a network managing audit exposure across hundreds of providers.

The Bottom Line for 2026

The AI scribe market for physical therapy has matured past the question of whether ambient documentation works. It works. The real dividing line in 2026 is architectural: whether the AI is a feature added to an existing system, or whether it's built into the operating system a clinic runs on — and whether that distinction has been tested at the scale of a real multi-site network, not just a single provider's workflow.

For growing rehab organizations evaluating this decision across multiple locations, the questions worth asking aren't about transcription quality alone. They're about whether the documentation layer can carry a note all the way to a clean claim, whether migration can happen without months of disruption, and whether the outcomes are backed by named, verifiable results rather than a vendor's general claims.

Related reading: SPRY vs. WebPT AI Scribe · SPRY vs. Prompt (Sidekick) AI Scribe · SPRY vs. Ensora AI Documentation · How AI Scribes Handle the Medicare 8-Minute Rule · KX Modifier & Therapy Threshold: Documenting It With AI · What 'Agentic' Documentation Means for Rehab Therapy

Frequently Asked Questions

Can I use an AI scribe for physical therapy on my phone?

Yes — most ambient AI scribes, including PT-specific and general-purpose tools, run through a phone or tablet's microphone placed in the treatment area, since therapists' hands are occupied during manual therapy and exercise instruction. For a single provider, this is usually as simple as an app. For a multi-site network, it raises governance questions — HIPAA/BAA coverage per device, consistent audio retention policy, and whether the note lands in the shared EMR or a personal app — that a native, EMR-embedded scribe answers consistently across every location by design.

What is the difference between a general AI medical scribe and a physical therapy–specific AI scribe?

A general AI medical scribe is typically trained on physician encounter structures — chief complaint, history, exam, assessment and plan. A physical therapy–specific scribe is built to capture range of motion in degrees, manual muscle testing grades, special test results, functional mobility status, and treatment time tied to timed CPT codes, none of which map cleanly onto a physician-style note.

Does an AI scribe handle Medicare's 8-minute rule automatically?

Some do. Platforms with native billing integration can capture treatment times from the encounter and calculate billable units according to the 8-minute rule thresholds. Documentation-only tools without a billing connection typically leave unit calculation to the provider or biller.

What is the KX modifier and why does it matter for AI documentation?

The KX modifier certifies that therapy services remain medically necessary once a patient's care crosses the Medicare therapy threshold ($2,330 for PT/SLP combined in 2026). Because it's a common audit target, documentation needs to clearly support functional progress and continued necessity — which is easier to guarantee consistently when the compliance check is built into the documentation workflow itself, rather than handled manually per provider.

Is ambient AI scribing accurate enough for physical therapy's specialized vocabulary?

Purpose-built PT AI scribes are trained to recognize rehab-specific terminology and structure — goniometric measurements, MMT grading (0–5 scale with modifiers), and standardized outcome measures like the Oswestry Disability Index, DASH, or Lower Extremity Functional Scale. General-purpose scribes not trained on this vocabulary are more likely to flatten precise measurements into vague narrative language.

What should a multi-clinic group prioritize differently from a solo practice when choosing an AI scribe?

Multi-site groups should weigh native billing integration, centralized compliance reporting, migration timelines, and role-based governance far more heavily than a solo practice would — because a documentation gap or billing inconsistency that's a minor annoyance for one provider becomes a scaled financial and audit risk across dozens of locations.

What's next for AI documentation in physical therapy after ambient scribing?

The category is moving toward agentic documentation — systems that retain context across visits, reason about clinical consistency, and take action (flagging issues, initiating billing workflows) rather than only transcribing. This shift is already visible in 2026 platform releases and is expected to become the baseline expectation for enterprise-grade systems within the next product cycle.

‍