Purpose/Significance To systematically review multi-graph convolutional network-based methods for traditional Chinese medicine (TCM) prescription recommendation, and to provide references for future research. Method/Process Literatures are retrieved to analyze the core principles, modeling mechanisms and performance of TCM prescription recommendation algorithms based on multi-graph convolutional network, and their limitations are discussed from the aspects of technical implementation and theoretical application. Result/Conclusion Future research should further strengthen the information fusion and relational modeling of multi-graph structures, deepen the exploration of interaction mechanism among symptoms, syndromes and herbs, and enhance the semantic representation and clinical interpretability of the models.