Mapping AI-supported formative assessment in physical education: a scoping review of process evidence and health-promotion relevance

Physical education (PE) provides opportunities for physical activity and health-related learning. Formative assessment may support both by enabling teachers and students to interpret and use evidence to guide teaching and learning. AI, learning analytics, and automated feedback may expand access to information on movement performance and learning processes. However, how AI-generated assessment information is used formatively and relates to health promotion remains unclear. This review mapped empirical research on AI-supported formative assessment in PE, examining process evidence, structural conditions, and outcomes within a potential health-promotion pathway. This scoping review followed PRISMA-ScR guidance. Five databases were searched from inception to 10 March 2026, with supplementary sensitivity searches conducted after refinement of the operational definitions. Eligible studies focused on PE or health and physical education, used an identifiable AI or intelligent analytical mechanism, and linked assessment evidence to formative purposes. Evidence was mapped using a “5 + 2” framework comprising five process-evidence dimensions, two structural conditions, and five levels of health-promotion relevance. Fifteen studies were included. Fourteen (93.3%) involved East or Southeast Asian settings, including 13 (86.7%) in mainland China or Taiwan. Evidence focused mainly on quantifiable movement or skill execution and AI-supported feedback generation and delivery. Engagement, learning progress, reflection, and self-regulation were examined, although less consistently as primary assessment focuses. Feedback interaction, teacher interpretation, data governance, equity, and sustainable implementation received limited attention. Reported outcomes were concentrated on proximal PE learning processes and potential health-promotion-related mediators. The included studies provided limited evidence on transfer beyond PE, sustained physical activity, health literacy, equity-related differences, and longer-term health outcomes. The review identified emerging formative uses of AI-generated assessment information in PE. However, the evidence was insufficient to establish continuity from classroom assessment to sustained physical activity or longer-term health outcomes. The findings emphasise the need to distinguish assessment information generation and delivery from its interpretation and use by students and teachers. Future research should examine feedback uptake, teacher judgement, maintenance over time, transfer across settings, implementation conditions, data governance, and the distribution of benefits and disadvantages across student groups.