JMIR Publications

ABSTRACT

Background:

Telerehabilitation improves access to care; however, without adequate human supervision, patients often struggle to adhere to prescribed programs. Motivation is a critical factor influencing rehabilitation outcomes, yet sustaining engagement in remote settings remains challenging. Recent advances in artificial intelligence offer new opportunities to personalize support and enhance patient participation.

Objective:

This scoping review examines current developments in AI-driven rehabilitation systems and highlights emerging trends and challenges, with a focus on the techniques employed, the evaluation methods used, and the user experiences reported across studies.

Methods:

Following the PRISMA-ScR guidelines, the review was conducted in accordance with established scoping review procedures. A total of 2,624 articles were retrieved from PubMed, IEEE Xplore, and Google Scholar databases (2000-2025). Of these, 138 full-text studies were assessed for eligibility, and 53 met the inclusion criteria. Extracted data were analyzed to identify the types of AI techniques deployed, evaluation methods applied, evidence of effectiveness, including comparisons with traditional rehabilitation when available, and reported challenges and facilitators experienced by patients and healthcare professionals.

Results:

Across the 53 included studies, a range of AI-driven techniques were deployed to support self-directed rehabilitation. Machine learning methods were the most common (34/53 studies), followed by natural language processing (13/53) and computer vision (12/53). Effectiveness was typically assessed using subjective and/or objective measures, such as self-reported motivation, usage intentions, program adherence, and task completion metrics. In studies that compared AI-driven tools with traditional rehabilitation, AI systems generally demonstrated higher engagement and adherence, although findings were mixed. Frequently reported challenges included technical issues, usability concerns, and variability in patient responses, while key facilitators included personalized feedback, gamified elements, and remote monitoring capabilities.

Conclusions:

Although motivation was not the primary focus in many of the reviewed studies, those that examined it reported promising outcomes. Overall, AI-driven approaches demonstrated positive effects on patient engagement and adherence, with conversational agents emerging as particularly promising tools. However, considerable heterogeneity in evaluation methods and a limited number of studies explicitly measuring motivational outcomes underscore the need for more systematic investigation. Future research should prioritize the design, integration, and evaluation of motivational features within AI-assisted rehabilitation systems to better understand their role in supporting sustained participation.