Abstract: Artificial intelligence (AI) has given rise to new ways of producing sports knowledge, driving the transformation of the scientific paradigm in sports. This study employs research methods such as literature review and logical analysis to explore the mechanisms, innovative forms, and practical pathways of AI-driven paradigm shifts in sports science. The new paradigm is based on autonomous learning from global corpora, triggering unrestricted collective intelligence emergence and the reconstruction of sports knowledge. The production of sports knowledge under this new paradigm involves stages such as the representation of sports behavior science, the aggregation of multi-source and multi-modal sports data, the generation of cross-disciplinary linked knowledge, and the optimization and iteration of sports knowledge. In terms of methodological innovation, quantitative and qualitative research methods in sports science are comprehensively optimized. Theoretically, by capturing deep inter-conceptual connections and overcoming the limitations of symbolic representation, it promotes traditional, modern, and global expressions in sports, Promoting the Construction of an Autonomous Knowledge System in Sports Science. Practically, it drives the transformation of sports research organizations from workshop models to open-source platforms, providing scientific theoretical support for the construction of a sports governance system through accurate, objective, and comprehensive knowledge output. To leverage the driving effects of the new paradigm, three measures are proposed: First, strengthen the development of sports corpora and gradually improve the distributed sports technology innovation system; second, enhance the digital literacy of sports practitioners to comprehensively improve the quality of digital talent training; third, implement coordinated governance of risks such as deepfakes, research misconduct, and intellectual property disputes, and improve the sports technology ethics governance system.