AI in Physical Therapy: What's Real in 2026, and What's Still Early

Key points

• AI in physical therapy is mostly used in three places: measuring movement from video, supporting home exercise, and reducing documentation time.
• The research base is growing but immature. A 2025 mapping review found 240 studies of AI in rehabilitation, but only 5.8% used external validation.
• The strongest evidence for clinical measurement is for markerless gait timing and distance measures. Joint angles outside the sagittal plane are less accurate.
• APTA supports AI that reduces administrative burden and enhances practice, with the physical therapist responsible for all aspects of patient management.

What the research landscape looks like

Two recent reviews give a fair picture of where the field is:

• A living mapping review of AI in rehabilitation from 2014 to 2024 identified 240 studies. Most focused on neurological (57.9%) and orthopedic (22.7%) rehabilitation, and most used supervised machine learning. External validation (5.8%) and explainability (10.2%) were rarely applied. The authors concluded the evidence still rests largely on small, single-type datasets (Morone et al., 2025).
• A systematic review of AI in physical rehabilitation screened 9,054 articles and included 28 projects, grouped into app-based systems, robotic devices, gaming systems and wearables. Across five randomized trials, clinical effects were inconsistent; barriers included technology literacy, reliability and user fatigue (Sumner et al., 2023).

So "AI in PT" is not one thing, and the evidence varies a lot by application. Here is where it stands for the three uses clinicians are most likely to meet.

1. Measuring movement from video

This is the most mature clinical use. Pose-estimation models find body landmarks in ordinary video and turn them into metrics such as gait speed, step length, joint angles and sit-to-stand time, without markers or sensors.

• A scoping review found markerless spatiotemporal gait measures generally appear equivalent to marker-based motion capture, while joint center locations and joint angles were not yet accurate enough for every clinical application (Wade et al., 2022).
• A meta-analysis of 22 studies of markerless 3D camera systems found good-to-excellent agreement for walking speed, step time and step length (ICC 0.81–0.98). Hip and knee angles in the sagittal plane were the most valid and reliable; ankle and out-of-plane measures were not (Scataglini et al., 2024).

The practical takeaway: trust AI video analysis most for timing, speed, distance and sagittal-plane angles, and be more cautious with frontal and transverse measures. Our guide to AI movement analysis for physical therapy covers how it works, and the evidence and validation page summarizes smartphone-specific studies.

2. Home exercise and adherence

AI can follow a patient's movement through the phone camera during a home program, counting repetitions and, in some products, giving feedback on form. The aim is to make between-visit work visible instead of self-reported. The evidence on whether digital tools improve adherence is mixed. We go through it in AI rep counting for home exercise programs.

3. Documentation

Ambient AI scribes listen to the visit and draft the note. The best trial evidence so far is outside rehabilitation: in a randomized trial of 238 outpatient physicians across 14 specialties, one of two scribes reduced time spent in notes by 9.5%, the other showed no significant change, and the authors noted that occasional inaccuracies require ongoing vigilance (Lukac et al., 2025). We found no peer-reviewed trial in a physical therapy setting yet. APTA has published a member practice advisory on AI-enabled ambient scribes covering documentation responsibilities and regulatory considerations.

Objective measurement helps documentation in a different way: numbers captured during a test go into the note directly, as we discuss in automated movement analysis and clinician burnout.

What APTA says

APTA's House of Delegates position "supports the ethical development and integration of artificial intelligence that reduces administrative burden and enhances physical therapist practice, education, and research." A companion position on digital health technologies and AI says they have the potential to expand access, enhance care delivery, promote safety, reduce administrative burden and improve outcomes, and lists criteria including that the physical therapist is responsible for all aspects of patient and client management.

How to evaluate an AI tool for your practice

• What exactly does it measure or do? "AI-powered" covers everything from a validated measurement to a chatbot.
• What's the evidence, for which outputs, against what reference? Look for validation of the specific metrics you'll use, ideally in patients like yours.
• Does it follow the standard test protocols your documentation and payers expect?
• Who stays in charge? The tool should support your decisions and make its outputs reviewable.
• Privacy and data: HIPAA alignment, a business associate agreement, and where video is processed and stored.

Our buyer's guide to movement analysis apps compares options by category.

Where Kinetically fits

Kinetically uses AI in the first two areas: it scores standardized movement tests from smartphone video (gait speed, Timed Up and Go, sit-to-stand, balance and sway, range of motion) and counts home-exercise reps with the phone camera, so clinicians see adherence and progress over time. It is decision support for the clinician, not a substitute for clinical judgment or diagnosis.

References

1. Morone G, Carbonetti R, Martino Cinnera A, et al. Artificial intelligence in rehabilitation: a living systematic mapping review – first release. Eur J Phys Rehabil Med. 2025;61(5):694-708. doi:10.23736/S1973-9087.25.09331-1
2. Sumner J, Lim HW, Chong LS, Bundele A, Mukhopadhyay A, Kayambu G. Artificial intelligence in physical rehabilitation: a systematic review. Artif Intell Med. 2023;146:102693. doi:10.1016/j.artmed.2023.102693
3. Wade L, Needham L, McGuigan P, Bilzon J. Applications and limitations of current markerless motion capture methods for clinical gait biomechanics. PeerJ. 2022;10:e12995. doi:10.7717/peerj.12995
4. Scataglini S, Abts E, Van Bocxlaer C, et al. Accuracy, validity, and reliability of markerless camera-based 3D motion capture systems versus marker-based 3D motion capture systems in gait analysis: a systematic review and meta-analysis. Sensors. 2024;24(11):3686. doi:10.3390/s24113686
5. Lukac PJ, Turner W, Vangala S, et al. Ambient AI scribes in clinical practice: a randomized trial. NEJM AI. 2025;2(12). doi:10.1056/AIoa2501000
6. American Physical Therapy Association. Ethical and effective integration of artificial intelligence across physical therapist practice, education and research. HOD P07-24-10-09. apta.org
7. American Physical Therapy Association. Digital health technologies, digital therapeutics, and artificial intelligence in physical therapist practice. HOD P07-24-12-11. apta.org
8. American Physical Therapy Association. APTA practice advisory on AI-enabled ambient scribe technology now available. September 19, 2025. apta.org