The study aims to investigate the effects of curcumin on radiation/chemotherapy-induced oral mucositis (R/CIOM) and preliminarily explore its mechanism. Randomized controlled trials were identified from the PubMed, Embase, Web of Science, Cochrane Library, Medline, and Google Scholar databases. RevMan 5.4 was used for statistical analysis to calculate the combined risk ratios (RRs). The mechanism was analyzed through network pharmacology, molecular docking, and a molecular dynamics simulation. The targets of curcumin were collected in HERB, PharmMapper, Targetnet, Swiss Target Prediction, and SuperPred. OMIM, GeneCards, and Disgenet were used to collect relevant targets for R/CIOM. Cytoscape software 3.8.0 was used to construct the component-target-pathway network. Protein–Protein Interaction (PPI) networks were constructed using the STRING database. GO and KEGG enrichment analyses were performed by Metascape. AutoDock Vina 4.2 software was used for molecular docking. The molecular dynamics simulation was performed by Gromacs v2022.03. It is found that 12 studies involving 565 patients were included. Meta-analyses showed that curcumin reduced the incidence of severe R/CIOM (RR 0.42 [0.24, 0.75]) and the mean severity of R/CIOM (MD -0.93 [−1.34, −0.52]). Eleven core target genes were identified in the treatment of R/CIOM with curcumin. The results of molecular docking and the molecular dynamics simulation showed that curcumin had strong binding energy and stability with target proteins including MAPK3, SRC, and TNF. Overall, these findings suggest curcumin can effectively improve severe R/CIOM, perhaps by affecting MAPK3, SRC, and TNF.