Abstract Background Aberrant DNA hypermethylation of CpG islands (CGIs) occurs frequently and is genome-wide in human gastric cancer (GC). A DNA methylation approach in plasma cell-free DNA (cfDNA) is attractive for the noninvasive detection of GC. Here, we performed genome-scale cfDNA methylation analysis in patients with GC. Methods We used MCTA-Seq, a genome-scale DNA methylation analysis method, on the plasma samples of patients with GC (n = 89) and control participants (n = 82), as well as 28 pairs of GC and adjacent noncancerous tissues. The capacity of the method for detecting GC and discriminating GC from colorectal cancer (CRC) and hepatocellular carcinoma (HCC) was assessed. Results We identified 153 cfDNA methylation biomarkers, including DOCK10, CABIN1, and KCNQ5, for detecting GC in blood. A panel of these biomarkers gave a sensitivity of 44%, 59%, 78%, and 100% for stage I, II, III, and IV tumors, respectively, at a specificity of 92%. CpG island methylation phenotype (CIMP) tumors and NON-CIMP tumors could be distinguished and detected effectively. We also identified several hundreds of cfDNA biomarkers differentially methylated between GC, CRC, and HCC, and showed that MCTA-Seq can discriminate early-stage GC, CRC, and HCC in blood by using a high specificity (approximately 100%) algorithm. Conclusions Our comprehensive analyses provided valuable data on cfDNA methylation biomarkers of GC and showed the promise of cfDNA methylation for the blood-based noninvasive detection of GC.