Abstract: Abstract: In the context of the rapid evolution of artificial intelligence (AI), interdisciplinary integration has become an increasingly vital pathway for driving technological breakthroughs and knowledge innovation. However, previous research rarely explores the interdisciplinary knowledge structure and its evolutionary dynamics in AI, limiting our understanding of its underlying development patterns and weakening the strategic initiative in shaping its trajectory. This study investigates the interdisciplinary knowledge integration and structural evolution within the AI domain. By introducing the BERTopic model and incorporating disciplinary diversity and cohesion indices as well as topic intensity and content evolution metrics, we develop a three-pronged analytical framework comprising “topic identification – structural analysis – mechanism comparison”. We identify representative interdisciplinary topics in the global AI landscape, systematically analyze their structural typologies and evolutionary trajectories, and further compare the evolutionary pathways and mechanisms of China and the United States. Our findings reveal 23 interdisciplinary topics in AI with distinct structural types and varying levels of knowledge integration, exhibiting characteristics of multi-paradigm coexistence, heterogeneous knowledge convergence, and structurally coupled evolution. This indicates that AI has emerged as a paradigmatic “interdisciplinary integration platform” within complex socio-technical systems. Moreover, China and the United States demonstrate two contrasting evolutionary paths: a “paradigm-driven, system-diffusion” model in the U.S., and a “task-oriented, technology-integration” model in China. These reflect fundamental institutional, organizational, and resource allocation differences—where the U.S. emphasizes open-ended exploration and pluralistic collaboration, and China prioritizes strategic orientation and mission-driven initiatives. Based on these insights, the study proposes policy recommendations such as building heterogeneous integration platforms, refining project funding mechanisms, and enhancing interface disciplines to strengthen China’s interdisciplinary capacity and systemic resilience in AI innovation. This research contributes to a deeper understanding of interdisciplinary knowledge production and evolution in AI and offers theoretical perspectives for evaluating the dynamics of Sino-US competition in emerging technologies.