Generative Artificial Intelligence and Large Language Models in Didactics of Sports Sciences and Physical Education: A Comprehensive Review of Pedagogical Applications, Teaching Innovations, and Research Implications

Abstract

Objective: The global burden of physical inactivity contributes to 5.3 million deaths annually, exceeding smoking-related mortality in certain regions and contributing substantially to the 1.5 billion individuals worldwide living with chronic diseases. Physical education (PE) represents a critical intervention point, yet persistent challenges limit effectiveness including inadequate instructional time (80 minutes weekly versus recommended 150 minutes), insufficient specialized teacher preparation (42% of elementary PE teachers lack specialized training), and limited capacity for differentiated instruction in heterogeneous student populations. The emergence of Generative Artificial Intelligence (Gen AI), particularly Large Language Models (LLMs) such as ChatGPT (launched November 2022, achieving 100 million users within two months), presents unprecedented opportunities for transforming pedagogical practices in sports sciences and PE contexts while simultaneously introducing critical challenges regarding academic integrity, cognitive development, and the preservation of embodied learning central to movement education.  This comprehensive review aimed to: (i) systematically examine current applications and pedagogical affordances of Gen AI and LLMs in sports sciences and PE didactics; (ii) analyze alignment with established pedagogical principles including constructivism, social constructivism, situated learning theory, and Universal Design for Learning; (iii) critically evaluate potential benefits and risks from a didactics perspective including impacts on teacher development, student learning outcomes, curriculum design, and assessment practices; and (iv) propose evidence-informed frameworks for pedagogically sound integration emphasizing human-AI collaboration rather than replacement of essential teaching functions.

Methods: A systematic literature search was conducted following adapted PRISMA guidelines across seven databases (PubMed/MEDLINE, Web of Science, Scopus, IEEE Xplore, ERIC, SPORTDiscus, Google Scholar) covering January 2022 to November 2025. Search terms combined Gen AI/LLM terminology with pedagogical concepts in sports sciences contexts using Boolean operators. Inclusion criteria focused on peer-reviewed articles examining pedagogical applications, teaching innovations, learning outcomes, and didactic research methodologies. From 1,247 initial records, 858 titles and abstracts were screened after duplicate removal (n=389), with 247 undergoing full-text review. Final analysis included 78 studies meeting inclusion criteria. Data extraction utilized standardized forms capturing study characteristics, methodological approaches, AI technology examined, pedagogical context, theoretical framework, key findings, and practice implications. Thematic analysis employed a pedagogically-oriented framework organizing findings into four domains aligned with core didactic functions: teaching support, learning enhancement, assessment innovation, and didactic research.

Results: Analysis of 78 studies (67% published 2023-2024) revealed significant pedagogical applications across four domains with concurrent identification of critical challenges. Teaching Support domain demonstrated lesson planning time reductions of 35% to 45%, with AI-assisted lesson plan quality rated 7.3 out of 10 (SD=1.2) for curriculum alignment compared to 7.8 out of 10 (SD=0.9) for manually created plans. Educators reported 67% satisfaction with differentiated instruction materials generated through Gen AI platforms. Learning Enhancement domain revealed improved conceptual understanding when engaging with AI tutors for anatomy and biomechanics concepts, with students valuing immediate availability and scaffolded explanations. Assessment Innovation applications showed AI-generated feedback demonstrated substantial agreement with expert teacher feedback for student assignments. Didactic Research efficiency gains included literature synthesis completion substantially faster than traditional methods and qualitative coding demonstrating substantial agreement between AI and expert human coders. However, critical challenges emerged including academic integrity violations in 23% to 43% of student work, factual inaccuracies in AI-generated specialized content, cognitive atrophy concerns (AI-Chatbot Induced Cognitive Atrophy, AICICA), reduced emphasis on embodied learning in PE contexts, and equity issues affecting students with limited digital literacy or technology access.

Conclusion: Gen AI and LLMs represent transformative tools for sports sciences and PE didactics when implemented within robust pedagogical frameworks that preserve the essential embodied, social, and affective dimensions of movement education. Evidence supports specific applications including administrative efficiency, differentiated cognitive content delivery, formative assessment support, and research methodology enhancement, while simultaneously demanding critical attention to academic integrity, accuracy verification, equity considerations, and prevention of cognitive atrophy through over-reliance. A hybrid pedagogical model is recommended integrating Gen AI for cognitive content delivery, theoretical knowledge construction, and administrative tasks while rigorously preserving face-to-face instruction, kinesthetic learning experiences, immediate physical feedback, and human mentorship central to sports education. Successful integration requires comprehensive teacher professional development focusing on pedagogical decision-making rather than technical operation, explicit policies balancing academic integrity with beneficial use, critical AI literacy curriculum specific to sports sciences contexts, and ongoing empirical evaluation of long-term learning outcomes. The field requires a paradigm shift from technology-driven adoption to pedagogy-informed integration, ensuring Gen AI serves educational goals rather than dictating them.