<p>A method for analyzing tumor evolution based on bulk RNA-sequencing data has not been reported yet. The epithelial–mesenchymal transition (EMT) is an evolutionarily conserved cellular program with high heterogeneity and plasticity. In this study, we proposed an EMT heterogeneity-based molecular typing (EHBMT) method to visualize cancer evolution and guide personalized medicine. Multiplex immunohistochemical assay and single-cell analysis were performed to confirm the feasibility of this method. EHBMT divided gastric (cancer) tissues into an epithelial phenotype cluster (EPC), hybrid epithelial–mesenchymal phenotype cluster (HPC) and mesenchymal phenotype cluster (MPC). Patients with gastric cancer with different EHBMT subtypes possessed distinct clinical features, molecular characteristics and prognostic outcomes. Furthermore, the proliferation ability of EPC, HPC and MPC subtypes decreases sequentially. Gene Ontology/Kyoto Encyclopedia of Genes and Genomes analysis showed that HPC subtypes are associated with inflammation and immune activation. More importantly, EHBMT discovered a sharp increase in the proportion of the HPC subtype during gastric cancer evolution. Traceability analysis indicated that the surge in HPC in gastric cancer was due to the transition from approximately 70–80% of normal EPC cases to cancerous HPC/MPC cases. In addition, the inflammatory factor IL-1β, highly expressed epithelial cells in the HPC subtype, should be a key driver for the decrease of epithelial cells by inducing EMT signaling. In conclusion, EHBMT is a novel method for visualizing cancer evolution using bulk transcriptomics. Gastric carcinogenesis is accompanied by a sharp increase in the proportion of HPC due to the abnormal EMT signaling pathway driven by an inflammatory microenvironment.</p>