Key points are not available for this paper at this time.
Aim: Artificial intelligence (AI) is frequently presented as a promising innovation in musculoskeletal rehabilitation. This systematic review aimed to identify clinical interventions described as “AI-assisted” in shoulder rehabilitation and to evaluate whether, based on their reported characteristics, they fulfilled predefined conceptual criteria for AI-supported therapy.Methods: Electronic searches were performed in PubMed/MEDLINE, Web of Science, Scopus, and CENTRAL from January 2010 to December 2025. Clinical studies evaluating digitally assisted rehabilitation for shoulder disorders were included. Each intervention was assessed using a predefined conceptual framework to determine whether it incorporated adaptive or trainable algorithmic components that directly informed rehabilitation delivery.Results: Fourteen eligible studies were identified. The investigated technologies mainly consisted of immersive virtual reality programs, gamified exercise platforms, wearable sensor feedback systems, and robotic or telerehabilitation devices. Most of the interventions operated using pre-defined or clinician guided exercise protocols. Based on the available descriptions, none clearly reported autonomous machine-learning algorithms capable of real-time adaptive treatment modification or independent clinical decision-making.Conclusion: Although AI terminology is increasingly used in the rehabilitation literature, currently reported shoulder rehabilitation interventions appear to rely predominantly on digital, virtual reality-based, sensor assisted or robotic systems rather than clearly described adaptive AI-driven therapeutic platforms. Based on published intervention descriptions, clinically implemented autonomous AI-supported rehabilitation appears to remain limited.
Dandinoğlu et al. (2026) studied this question.