Abstract: This study applies an integrated approach that combines the BERTopic model with Large Language Models(LLMs) to systematically analyze 1160 domestic research articles on AI literacy. Twenty core research topics are identified and further categorized into three major research directions: the integration of intelligent technologies with educational and academic frontiers; multidimensional integration driven by policy and societal needs; and context-based practices that deepen the integration of education and intelligent technologies. The findings suggest that AI literacy research in China has gradually formed a “technology-practice-institution-theory” ecosystem, while notable challenges remain, including imbalanced research orientations, underdeveloped evaluation frameworks, and the absence of long-term perspectives. Building on these insights, the study highlights the role of university libraries as essential hubs connecting technology, education, and society, and proposes three practical pathways: constructing an inclusive AI literacy education ecosystem, developing a localized evaluation framework, and advancing long-term scenario-based educational programs. These recommendations aim to contribute to both the theoretical enrichment and practical innovation of AI literacy education.