Background: Gastric cancer (GC) is one of the most fatal cancers in the world. Results of previous studies on the association of the CpG island methylator phenotype (CIMP) with GC prognosis are conflicting and mainly based on selected CIMP markers. The current study attempted to comprehensively assess the association between CIMP status and GC survival and to develop a CIMP-related prognostic gene signature of GC. Methods: We used a hierarchical clustering method based on 2,082 GC-related methylation sites to stratify GC patients from the cancer genome atlas into three different CIMP subgroups according to the CIMP status. Gene set enrichment analysis, tumor-infiltrating immune cells, and DNA somatic mutations analysis were conducted to reveal the genomic characteristics in different CIMP-related patients. Cox regression analysis and the least absolute shrinkage and selection operator were performed to develop a CIMP-related prognostic signature. Analyses involving a time-dependent receiver operating characteristic (ROC) curve and calibration plot were adopted to assess the performance of the prognostic signature. Results: We found a positive relationship between CIMP and prognosis in GC. Gene set enrichment analysis indicated that cancer-progression-related pathways were enriched in the CIMP-L group. High abundances of CD8+ T cells and MI macrophages were found in the CIMP-H group, meanwhile more plasma cells, regulatory T cells and CD4+ memory resting T cells were detected in the CIMP-L group. The CIMP-H group showed higher tumor mutation burden, more microsatellite instability-H, less lymph node metastasis, and more somatic mutations favoring survival. We then established a CIMP-related prognostic gene signature comprising six genes (CST6, SLC7A2, RAB3B, IGFBP1, VSTM2L and EVX2). The signature was capable of classifying patients into high-and low-risk groups with significant difference in overall survival (OS; p < 0.0001). To assess performance of the prognostic signature, the area under the ROC curve (AUC) for OS was calculated as 0.664 at 1 year, 0.704 at 3 years and 0.667 at 5 years. When compared with previously published gene-based signatures, our CIMP-related signature was comparable or better at predicting prognosis. A multivariate Cox regression analysis indicated the CIMP-related prognostic gene signature was an independent prognostic indicator of GC. In addition, Gene ontology analysis indicated that keratinocyte differentiation and epidermis development were enriched in the high-risk group. Conclusion: Collectively, we described a positive association between CIMP status and prognosis in GC and proposed a CIMP-related gene signature as a promising prognostic biomarker for GC.