Artificial intelligence in physical and rehabilitation medicine: first update of the living systematic mapping review up to May 31st, 2025.

BACKGROUND: Artificial intelligence is increasingly applied to rehabilitation; however, persisting gaps in research, such as methodological heterogeneity, poor reporting, and lack of validation, reduce clinical translation. Living systematic reviews offer a dynamic approach to continuously monitor evolving evidence. The aim of this first update up to May 31st, 2025 is to expand current evidence base, identifying emerging trends and persistent gaps. METHODS: A total of 784 studies were screened, and 95 new studies were included, for a cumulative total of 335 studies across the first release and the update up to May 31st, 2025. RESULTS: In this update up to May 31st, 2025, 95 new studies were included, bringing the cumulative dataset to 335 studies. A strong recent growth in publications was observed, with China and the USA as the main contributors. There is an increase in the clinical-oriented use of AI. Of note, there is a slight increase in the adoption of reporting guidelines. Methodological limitations persisted, particularly a lack of adoption of reporting tools dedicated to and validated for AI. Explainability remains inconsistently described and applied (28.4%) with a growth trend (+18.2%); and external validation is infrequent (13.7%). Although randomized controlled trials showed a slight increase, they remain underrepresented overall, a finding consistent with the OCEBM distribution, which is still predominantly weighted toward lower-to-moderate levels of evidence (Level 3-4). This highlights persistent gaps in high-quality evidence. CONCLUSIONS: Literature trends of the update of the living systematic mapping review REHALISE up to May 31st, 2025, compared to the first release, suggest a slight and gradual shift toward more clinically oriented applications, though robustness and reproducibility remain critical concerns.