Large language models linking traditional Chinese medicine knowledge and clinical practice

Traditional Chinese medicine (TCM) is a complex medical system characterized by multi-source data, implicit knowledge representation, and syndrome-based diagnostic and therapeutic reasoning. With the increasing digitization of classical texts and clinical records, large language models (LLMs) have attracted growing interest as computational tools for organizing and utilizing knowledge related to TCM. However, existing studies remain scattered, and a comprehensive overview of data resources, modeling strategies, evaluation practices, and application scenarios is still lacking.

In this review, we summarize current research on the application of LLMs in TCM based on published literature. We focus on commonly used data sources, including classical texts, clinical records, and related structured resources, and review representative modeling approaches such as knowledge-enhanced and multimodal methods. Domain-specific training and fine-tuning strategies, as well as reported evaluation practices, are also summarized. Furthermore, we review representative application scenarios described in the literature, including medical consultation support, syndrome differentiation assistance, prescription-related support, education, and research assistance. Key limitations, such as limited interpretability, safety concerns, and the lack of standardized evaluation frameworks, are discussed.

By organizing existing studies along the workflow from data construction to model development, evaluation, and application, this review aims to clarify the current research landscape and highlight methodological challenges that should be addressed to support cautious and appropriate use of LLMs in TCM research and practice.