ABSTRACT Traditional Chinese Medicine (TCM), with its fragmented knowledge system, subjective diagnostics, and limited standardization, faces ongoing challenges in modernization. The rapid development of Large Language Models (LLMs) offers an unprecedented opportunity for the digitalization and standardization of TCM knowledge. By leveraging deep semantic understanding, contextual reasoning, and knowledge‐graph–based inference, LLMs can systematize knowledge, integrate multimodal data, and support both standardized and personalized decision‐making. Several TCM–LLMs, such as Qibo and MedChatZH, have been developed. Although previous reviews compared individual TCM–LLMs, a comprehensive analysis of common characteristics is lacking. This review addresses this gap by examining the “construction–application–challenges–prospects” paradigm of TCM–LLMs. Construction generally involves collecting authoritative data from diverse sources, standardizing textual content, training and fine‐tuning on foundational LLMs, and subsequent evaluation. Key applications include auxiliary diagnosis and treatment, health management, medical education, and drug discovery. Five major challenges are data quality, reasoning performance, multimodal integration, ethical and regulatory compliance, and cultural adaptability. On the basis of this analysis, we propose a classification scheme that categorizes applications into four types and adopt the identified challenges as evaluation dimensions for future TCM–LLMs. This framework aims to clarify conceptual structures, guide methodology, and provide a foundation for innovation and cross‐cultural integration in TCM research.
Jiang et al. (Mon,) studied this question.