Breaking Boundaries: The Fusion of Physiotherapy and AI

TL;DR

Clinical and billing decisions still require clinician review and confirmation.

• Documentation tools draft SOAP notes based on visit conversations, reducing the manual charting burden discussed in APTA’s coverage of AI documentation tools.
• Exercise selection tools help clinicians retrieve relevant options. Physitrack’s AI Exercise Search provides one current example.
• Remote monitoring tools use computer vision and patient data to track movement quality and home exercise adherence.
• Administrative tools automate eligibility checks and flag claim denial risks before submission.

Separating what's shipping from what's hype in PT

AI in physical therapy already supports four defined areas. Documentation tools draft notes, exercise tools suggest relevant movements, remote platforms analyze functional data, and administrative systems assist with scheduling and billing. The American Physical Therapy Association has covered ambient AI scribes as a practical issue that clinicians are actively evaluating.

Each category automates a time-consuming task while leaving consequential decisions to a qualified human. A clinician reviews note text, confirms exercise choices, interprets monitoring alerts, and checks suggested billing codes. Products that claim to remove that oversight deserve greater scrutiny than tools that clearly define where review occurs.

Evidence quality also varies by category. General healthcare studies provide useful benchmarks for documentation and administrative automation, but PT-specific adoption and outcome data remain limited. Vendor figures can indicate possible time savings, engagement, or financial effects, but clinics should treat them as directional until independent studies verify the results in physical therapy settings.

Clinical documentation: the ambient scribe shift

A multicenter JAMA study found that ambient AI scribes saved clinicians 16 minutes per eight-hour shift, while Cleveland Clinic reported 14 fewer minutes of daily EHR work. A separate JAMA study recorded a 31 percent reduction in self-reported burnout among users. These cross-healthcare findings provide a useful baseline, but they do not establish equivalent savings in physical therapy clinics.

An ambient scribe captures the clinical conversation and turns it into a structured draft note. The software can organize patient history, examination findings, and treatment details while the clinician conducts the visit. The clinician then reviews the draft, corrects mistakes, and signs the final record.

Ambient scribe drafts still require close review. The 2026 multicenter JAMA study found that 15 to 20 percent of notes contained errors requiring correction, especially in the assessment and plan. Those sections often carry the clinical reasoning that guides treatment, so a plausible sentence can still create a meaningful error if the clinician accepts it without checking the source conversation.

AI-generated documentation can also affect coding. A policy brief in npj Digital Medicine warned that ambient scribes may push evaluation and management coding toward higher complexity levels than an encounter supports. That finding comes from broader healthcare settings rather than physical therapy specifically, but it shows why clinics should audit code suggestions against the visit record instead of treating them as final.

The American Physical Therapy Association frames ambient scribes as an active professional issue, with physical therapists weighing time savings against accuracy and other risks. PT-specific adoption rates and time-saving studies remain limited, so clinic owners should treat broader healthcare benchmarks as directional. A local pilot should measure editing time and correction frequency before a clinic expands use.

AI-assisted exercise selection and prescription

AI-assisted exercise tools help clinicians retrieve suitable options for a patient's specific treatment needs. Conventional library searches often depend on exact exercise names or filters, while AI search can interpret a clinical description and return likely matches for review.

More advanced tools can use movement and recovery data to refine those suggestions. Computer vision analyzes video from a phone or standard camera to measure movement patterns, range of motion, and exercise quality without wearable sensors. Predictive models can compare those observations with historical patient data to estimate recovery patterns or flag possible reinjury risk, according to a technical overview from Hemscap. A platform could use those signals to suggest an easier variation, a progression, or a different movement category.

Physitrack’s AI Exercise Search provides one current example of AI-assisted retrieval. The tool helps clinicians find relevant options within Physitrack’s exercise library without relying solely on exact keywords. Clinicians can review the results, select appropriate exercises, and set the dosage and progression within the patient’s program.

Current AI tools cannot independently account for every factor that shapes a safe prescription. Diagnosis, tissue irritability, precautions, and the patient’s response during assessment may change an otherwise reasonable selection. Computer vision can also misread movements when camera placement, lighting, or visibility limits the input. Clinicians therefore remain responsible for confirming each exercise, adapting the program, and monitoring how the patient responds.

Remote monitoring and data-driven rehab

Remote monitoring extends clinical observation between visits by collecting data on how a patient follows and responds to treatment. Remote patient monitoring, or RPM, tracks physiologic measures such as blood pressure or glucose. Remote therapeutic monitoring, or RTM, instead captures nonphysiologic information such as pain, exercise adherence, treatment response, and musculoskeletal status through qualifying software classified as a medical device.

The markets for both categories continue to grow. One industry estimate valued the global RPM market at $39.5 billion in 2023 and projected a value of $77.9 billion by 2029. The same estimate projected that the U.S. RTM segment would grow at 17.2 percent annually between 2025 and 2030, reaching $969.8 million by 2030. These forecasts indicate commercial adoption, but they do not prove that every monitoring program improves physical therapy outcomes.

AI can turn collected functional data into signals that a clinician can review. Computer vision tools can use a phone or computer camera to estimate range of motion and identify movement patterns. Predictive models can compare current and historical data to flag declining adherence, slower progress, or elevated reinjury risk. A physical therapist still needs to interpret those signals within the patient’s clinical context and decide whether to change the program.

Successful adoption requires training, privacy safeguards, and reliable access. Clinicians need to understand how a tool produces alerts and how often those alerts prove accurate. Clinics must assess how vendors store, transmit, and control access to protected health information. Patients also need compatible devices and dependable internet service, which can limit participation. Independent evidence specific to physical therapy remains limited, so clinics should ask vendors for validation in patient populations and workflows that resemble their own.

Administrative automation: scheduling, billing, and prior authorization

Administrative automation reduces repetitive data entry before and after each visit. Digital intake tools can capture demographics, insurance details, consent forms, and copays before check-in. Scheduling systems can support online booking, send reminders, and offer canceled appointments to people on a waitlist. Some platforms also use historical attendance data to estimate no-show risk, though clinics should check how accurately those predictions reflect their own patients.

Billing tools review documentation before a claim reaches the payer. Claim-scrubbing software can flag missing fields, inconsistent codes, and documentation that may not support medical necessity. Coding assistants suggest diagnosis or procedure codes, while claims platforms track payer responses and route denials for staff review. These tools can reduce preventable errors, but a qualified staff member still needs to confirm the code and supporting documentation.

Prior authorization automation is gaining attention as federal requirements change payer workflows. Under the CMS Interoperability and Prior Authorization Final Rule, affected payers must issue standard decisions within seven calendar days and expedited decisions within 72 hours beginning January 1, 2026. Payers must also provide specific denial reasons, and new API requirements take effect January 1, 2027. The administrative burden remains substantial. A 2024 AMA survey found that physician practices handled an average of 39 requests per physician each week, requiring about 13 staff hours. Although those figures are not specific to physical therapy, they explain the demand for tools that assemble documentation, submit requests, and track decisions.

Any vendor handling protected health information should meet a clear compliance baseline. A clinic should require a signed Business Associate Agreement and confirm that the vendor maintains the administrative, physical, and technical safeguards required under HIPAA. A current SOC 2 report or HITRUST certification can support due diligence, but neither replaces the clinic's own security review. Clinics should ask where the vendor stores PHI and whether it uses patient data to train models. They should also review access controls, deletion policies, and incident-response terms before connecting an automation tool to your records or billing systems.

How to evaluate an AI tool before adopting it

Evaluate an AI tool by testing its decisions, workflow fit, and safeguards. A polished interface reveals little about whether the tool can support daily clinical work.

• Ask the vendor to separate automated tasks from clinician-controlled decisions. Clinicians should confirm note text, exercise selection, responses to monitoring alerts, and billing codes before those outputs enter the record.
• Test accuracy with realistic cases. Review documentation errors, unsuitable exercise suggestions, missed monitoring alerts, and incorrect coding recommendations. Ask how the vendor measures errors and updates its models.
• Verify integration with your practice management or EHR system. The tool should reduce duplicate entry and preserve a clear record of patient activity, clinician review, and changes.
• Examine how the vendor handles protected health information. Request its business associate agreement, HIPAA safeguards, access controls, audit logs, data retention terms, and relevant independent security assessments.
• Run a limited pilot with defined measures. Track clinician time, correction rates, patient adherence, administrative workload, and staff support needs before expanding use.
• Confirm that the vendor provides training and a clear support path. Clinicians need to understand when an AI output requires extra scrutiny and how to report a problem.

Physitrack’s AI Exercise Search follows a human-confirmed model for exercise prescription. The tool helps clinicians find relevant exercises, while the clinician reviews each option and determines the patient’s program. Explore Physitrack’s exercise platform.

Closing takeaway

Clinics that adopt AI successfully will assign it narrow, measurable jobs and keep clinicians responsible for consequential decisions. They will test accuracy, review outputs, monitor errors, and confirm that each tool fits existing clinical and compliance workflows.

Hasty adoption starts with a broad promise and adds oversight after problems appear. Over the next few years, useful AI tools will earn their place by reducing specific burdens while leaving assessment, treatment planning, and billing confirmation with qualified people. AI works best when it supports staff under clear human supervision.

FAQs

Is AI replacing physical therapists?

AI can assist with defined tasks, but it cannot independently assess patients or accept clinical responsibility. With tools such as Physitrack, clinicians remain responsible for exercise suitability and safety.

How accurate are AI clinical notes?

AI notes require review because 15 to 20 percent of notes needed corrections in one multicenter study. Clinicians should apply that review standard to all AI-generated notes. Review is especially important in the assessment and plan sections.

How does RTM differ from RPM?

RTM tracks pain, adherence, and musculoskeletal status, while RPM tracks physiologic measures such as blood pressure. Clinics evaluating an RTM platform should confirm which monitoring and billing rules apply. The distinction determines eligible data and devices.

Does insurance cover AI-monitored care?

Insurance may cover qualifying RTM services, but payer rules, devices, and documentation determine payment. Clinics considering AI-monitored care should verify coverage with payers and compliance advisers. No AI tool can guarantee reimbursement.

What should a clinic check before buying an AI tool?

Clinics should check accuracy, review controls, integrations, data handling, and support. When evaluating Physitrack or another AI tool, clinics should establish exactly what clinicians must confirm. Vendors handling protected health information should sign a BAA and maintain HIPAA safeguards.