Rethinking ethical governance of generative AI in sport pedagogy research: a discipline-sensitive perspective

Generative artificial intelligence (GenAI)-including large language models, image synthesis, and learning analytics tools, is increasingly used in pedagogical research for instructional design, feedback generation, data analysis, and manuscript preparation. However, discipline-specific ethical guidance remains limited, particularly in fields characterized by embodied learning and performance-based pedagogy. In sport sciences, GenAI applications intersect with motor learning, sensorimotor feedback, and learner autonomy, raising distinctive challenges related to academic integrity, transparency, authorship, data protection, and pedagogical validity. In this Perspective, we argue that existing institutional and international AI ethics frameworks, while essential, remain predominantly text-oriented and insufficiently responsive to the embodied, safety-sensitive, and movement-based characteristics of sport pedagogy research. This creates a governance gap in which AI-generated feedback may directly affect motor learning processes, learner safety, and skill acquisition trajectories in ways not adequately captured by generic AI ethics guidance. Drawing on research integrity standards, global AI governance principles, and motor learning theory, we identify key ethical tensions emerging from the integration of generative systems into sport pedagogical research and practice. This article develops a discipline-sensitive ethical framework through a structured, purposive, and theory-guided synthesis of relevant literature. We propose a set of priority orientations to support responsible GenAI use in this context, including strengthened human accountability, transparent disclosure practices, discipline-informed validation of AI-generated feedback, and enhanced protection of learner data. These propositions require empirical validation in real world sport pedagogy contexts. We position pedagogical validity, defined as alignment between AI-supported processes and established motor learning principles, as a novel ethical criterion not yet explicitly articulated in AI education governance. Finally, we outline a research and policy agenda for empirically validating this framework across diverse embodied learning environments and clarify the logic used to identify and synthesize literature across AI governance, research integrity, and motor learning scholarship. This conceptual, theory-guided Perspective is intended to inform researchers, ethics committees, and higher education institutions seeking disciplines-sensitive approaches to the responsible integration of GenAI in embodied learning and performance-based education contexts.