EFFECTIVENESS OF ARTIFICIAL INTELLIGENCE–ASSISTED PHYSIOTHERAPY AND REHABILITATION IN ADULTS WITH MUSCULOSKELETAL DISORDERS: A SYSTEMATIC REVIEW

Background: Artificial intelligence (AI) is increasingly integrated into musculoskeletal rehabilitation through adaptive self-management, computer-vision exercise feedback, digital exercise prescription, and AI-supported postoperative care. Clinical trials are heterogeneous, and recent systematic reviews have produced different conclusions depending on the population and technology examined. Objective: To systematically synthesize randomized controlled trial evidence on AI-assisted physiotherapy and rehabilitation for adults with musculoskeletal disorders, focusing on pain, disability/function, mobility, adherence, patient experience, and healthcare utilization. Methods: An updated evidence search was conducted through 20 August 2026 using PubMed/PMC-indexed records and targeted searches of publisher and journal databases, supplemented by reference-list screening of recent systematic reviews and meta-analyses. Eligible studies were randomized controlled trials in adults with musculoskeletal disorders in which an AI component actively contributed to rehabilitation, exercise guidance, self-management, movement feedback, or postoperative mobility. Studies involving neurological populations, healthy volunteers, or digital interventions without a genuine AI component were excluded. Risk of bias was considered using RoB 2 domains. Because interventions and outcomes were heterogeneous, the primary synthesis was narrative. Results: Eight randomized controlled trials were identified as clinically relevant to the revised scope: Sandal 2021, Priebe 2020, Itoh 2022, Park 2023, Marcuzzi 2023, Xiao 2025, Alptekin and Ulubay 2026, and Villar-Alises et al. 2026. The studies covered low back/neck pain, rotator cuff-related shoulder pain, and postoperative total knee arthroplasty. Evidence was mixed. The selfBACK trial reported a small reduction in low-back-pain-related disability, while the later specialist-care SELFBACK trial found no significant benefit on its primary MSK-HQ outcome. A 2026 meta-analysis of five RCTs in nonspecific low back pain found statistically significant but clinically small pooled reductions in pain and disability, with low/very-low certainty. The 2026 shoulder and postoperative TKA trials reported favorable short-term outcomes, but both were small or context-specific. Conclusion: AI-assisted rehabilitation may provide useful adjunctive support for exercise delivery, movement feedback, self-management, and postoperative education/mobility. Current evidence does not support replacing physiotherapist assessment or treatment with AI. Effects are generally modest, technology-specific, and uncertain in the long term. Keywords: Artificial intelligence; physiotherapy; physical therapy; musculoskeletal rehabilitation; low back pain; shoulder pain; knee arthroplasty; telerehabilitation; computer vision; digital rehabilitation