Artificial intelligence in the delivery of physical and health education

This paper examined how artificial intelligence should be integrated into the delivery of Physical and Health Education (PHE), a discipline whose core content, unlike that of Mathematics or English, is embodied, practical and relational rather than purely informational. Physical and Health Education delivery is shaped by structural strain, large class sizes, inconsistent feedback, uneven engagement, and heavy teacher workloads, that AI is increasingly proposed to address. The paper asks not simply whether AI belongs in PHE, but which of the discipline's functions can be meaningfully augmented by AI, which should remain fundamentally human, and what conditions should govern the boundary between the two. AI should be strategically integrated into PHE as a pedagogical support tool rather than a replacement for educators, with its role bounded by ethical safeguards, equitable access, and preservation of the discipline's practical, embodied nature. The appropriateness of AI in PHE depends on the educational domain being served: it clusters around outcomes that are structured and measurable, and diminishes sharply around outcomes that are relational, moral, and social. The paper defined the relevant AI technologies before examining their current applications in PHE across eight categories, including personalized learning, motion analysis, wearable technology, virtual and augmented reality, and health education. It builds the case for integration around documented evidence of improved personalization, engagement, accessibility, and teacher support, then weighs this against genuine counterarguments: the digital divide's weight in infrastructure-constrained regions, teacher displacement fears, ethical risks around bias and surveillance, threats to academic integrity, loss of human interaction, and a definitional limit, whether AI can meaningfully teach teamwork or values, that current evidence does not resolve. Ethical and policy conditions for adoption are set out, alongside recommendations for governments, schools, teachers, curriculum developers, and researchers. AI's demonstrated strengths cluster around PHE's psychomotor and cognitive domains; its affective and social domains remain, and should remain, the province of the human teacher. Benefits and risks are not fully symmetrical: many risks are governance failures addressable through policy, while other risks, including persistent algorithmic bias and the behavioural effects of continuous monitoring, call for ongoing evaluation rather than a one-time fix. Current evidence does not establish that AI can substitute for the relational and social learning PHE's affective domain depends on, which is why those functions should not yet be delegated to AI. The paper's central contribution is this domain-specific standard for integration, rather than a categorical verdict for or against AI in PHE.