PARP-1 and MGMT play an important role in the DNA repair system and therefore have been implicated in human carcinogenesis. However, the association between the most studied PARP-1 rs1136410: T > C and MGMT rs12917: C > T polymorphism and risk of gastrointestinal (GI) cancers was reported with inconclusive results. Accordingly, a meta-analysis of 23 published case-control studies was conducted to assess the strength of association using crude odds ratios (ORs) with 95% confidence intervals (CIs). Overall, the C allele of PARP-1 rs1136410: T > C polymorphism was significantly associated with increased susceptibility of GI cancers (homozygote comparison: OR = 1.43, 95% CI 1.14-1.81; heterozygote comparison: OR = 1.18, 95% CI 1.07-1.29; dominant model: OR = 1.23, 95% CI 1.12-1.35; recessive model: OR = 1.30, 95% CI 1.04-1.62; allelic comparison: OR = 1.19, 95% CI 1.07-1.32). In the subgroup analysis, still obvious associations were found in the Asian population, gastric cancer, and high-quality studies. For MGMT rs12917: C > T polymorphism, no obvious associations were found for all genetic models overall. However, in the subgroup analysis, we found that the T allele was significantly associated with reduced colorectal cancer risk for heterozygote (OR = 0.83, 95% CI 0.70-0.97) and dominant model (OR = 0.84, 95% CI 0.72-0.98). In conclusion, this meta-analysis suggests that the PARP-1 rs1136410: T > C polymorphism is a susceptibility factor for GI cancers, but the variant allele of MGMT rs12917: C > T polymorphism appears to be a protective factor for colorectal cancer. Large-scale and well-designed case-control studies are necessary to validate the risk identified in the present meta-analysis.