Effects of AI-based real-time feedback on students’ motor-skill performance in physical education: a systematic review and meta-analysis

Introduction Timely feedback may improve motor-skill performance in physical education, but immediate gains do not establish durable motor learning without retention or transfer assessment. This review evaluated AI-based real-time feedback for students’ post-intervention motor-skill performance in educational settings. Methods Following PRISMA 2020, Web of Science, ERIC, PubMed, and CNKI were searched for journal articles published from 1 January 2014 to 25 June 2026. Eligible studies examined AI-based or intelligent feedback and motor-skill outcomes in educational learners using quantitative or mixed-methods designs. Risk of bias was appraised using design-specific JBI tools. Study-level Hedges’ g values were synthesized using a REML random-effects model, and certainty was assessed using GRADE. Results Eighteen studies were included in the systematic review and 11 contributed study-level effects to the meta-analysis. One retracted report and one purely qualitative study were excluded during verification. The pooled effect favored AI-based real-time feedback (Hedges’ g = 1.484, 95% CI [0.738, 2.230]). Heterogeneity was substantial (I 2 = 94.70%; τ 2 = 1.50748), and the 95% prediction interval was −1.036 to 4.003. The study with the largest effect was potentially influential; excluding it reduced the pooled effect to g = 1.175 (95% CI [0.679, 1.671]). Appraisal identified recurring concerns about blinding, allocation concealment, follow-up, and statistical reporting. Discussion AI-based real-time feedback showed a positive average association with post-intervention motor-skill performance, but extreme heterogeneity and a prediction interval crossing zero limit conclusions about effects in individual settings. None of the pooled outcomes assessed delayed retention or transfer, so durable motor learning cannot be inferred. GRADE certainty was very low due to serious risk of bias, very serious inconsistency, and serious indirectness. The review is registered in PROSPERO (CRD420261477139). Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261477139 , PROSPERO CRD420261477139.