Background: Exposure to environmental carcinogens can cause damages to DNA. If not properly repaired, the DNA damages may increase the risk of carcinogenesis. Xeroderma pigmentosum group G (XPG) gene is an essential gene in the nucleotide excision repair (NER) pathway. The association between XPG polymorphisms and cancer susceptibility has been the focus of attention in the molecular epidemiology of cancer. However, the conclusions have been divergent. Therefore, we conducted a comprehensive meta-analysis to precisely evaluate the association of 3 frequently investigated XPG polymorphisms (rs751402, rs873601, and rs2296147) with cancer risk. Methods: Pubmed, EMBASE, and Chinese National Knowledge Infrastructure (CNKI) were searched for relevant studies in English and Chinese. Odds ratio (OR) and 95% confidence interval (CI) were used to assess the association between XPG polymorphisms (rs751402, rs873601, and rs2296147) and cancer risk. Results: Twenty-three studies were included. Overall, there was no significant association between rs751402 polymorphism and overall cancer risk under the 5 gene models. However, we observed strong correlation between rs751402 polymorphism and gastric cancer (C vs T: OR=1.21, 95% CI = 1.00–1.26, P = .045; TC vs CC: OR = 1.12, 95% CI = 1.00–1.24, P = .041; TC/TT vs CC: OR = 1.13, 95% CI = 1.02–1.26, P = .020). There was a significant correlation between rs873601 polymorphism and cancer risk under the homozygous model (GG vs AA: OR = 1.16, 95% CI = 1.07–1.26, P = .001). Moreover, significant association with breast cancer was detected for rs873601 polymorphism under the allele contrast model (G vs A: OR = 1.10, 95% CI = 1.02–1.20, P = .021). In the subgroup of Asian, rs873601 polymorphism was related to the susceptibility to cancer (G vs A: OR = 1.07, 95% CI = 1.03–1.12, P = .010; GG vs AA: OR = 1.15, 95% CI = 1.06–1.26, P = .001; AG/AA vs GG: OR = 1.08, 95% CI = 1.01–1.15, P = .031; AA vs AG/GG: OR = 1.13, 95% CI = 1.05–1.21, P = .001). Significant association between rs2296147 polymorphism and cancer risk were observed in Asian population (CT vs TT: OR = 0.93, 95% CI = 0.87–0.99, P = .036). Conclusions: Our meta-analysis suggested that the rs873601 polymorphism was significantly associated with overall cancer risk. The moderate effects of rs751402 and rs2296147 polymorphism on cancer susceptibility might be highly dependent on cancer type and ethnicity, respectively. Large studies are needed to validate our findings, especially in Caucasian and African population.