Abstract: 【Purpose/significance】Since OpenAI released ChatGPT in late 2022, generative artificial intelligence has advanced rapidly in many fields. This study aims to clarify the structure and evolution of research themes in this academic area to reveal the dynamic patterns of key topics and future directions.【Method/process】Using the BERTopic model, we perform text semantic embedding, UMAP dimensionality reduction, and HDBSCAN density clustering. This approach, combined with dynamic topic analysis, accurately identifies research topics and maps their evolutionary paths.【Result/conclusion】Twenty specific research topics are identified and mapped, including intelligent education technology, risk and technology governance, and AI-powered content services. The results show that the research focus has gradually shifted from technological exploration to specific applications. Among these, intelligent edu⁃ cation has remained a core research area, while risk governance has gained increasing attention. Dynamic analysis reveal clear paths of integration and differentiation in topic evolution, reflecting both interdisciplinary integration and topic specialization.【Innovation/ limitation】The innovation in both technology and methodology improves the visibility and interpretability of the research results. How⁃ ever, the limited diversity of data sources may have led to incomplete coverage of certain research topics.