The RAD51 gene is essential for the repair of damaged DNA related to tumor development. Although a number of genetic studies have attempted to link the 135G/C polymorphism of RAD51 gene to the risk of cancer, the results were inconclusive. The present study aimed at investigating the pooled association using the more comprehensive meta-analysis. The PubMed, EBSCO, and BIOSIS databases were searched to identify eligible studies which were published in English before March 2014. Data were extracted using standardized methods. The association was assessed by odds ratio (OR) with 95 % confidence interval (CI). Begg's test was used to measure publication bias. Sensitivity analyses were also performed to assess the stability of the results. A total of 45 eligible studies with 28,956 patients and 28,372 controls were included in this meta-analysis. Overall, significant association was detected between 135G/C polymorphism and increased cancer risk (C allele vs. G allele: OR 1.23, 95 % CI 1.18-1.28; CC vs. GG: OR 2.41, 95 % CI 2.12-2.74; CC vs. CG: OR 3.86, 95 % CI 3.41-4.37; recessive model: OR 3.57, 95 % CI 3.19-4.00). In further stratified analysis, significantly elevated cancer risk was observed among Caucasians but not Asians. Subgroup analysis by different cancers also showed their significant associations in breast cancer, hematologic malignances, ovarian cancer, colorectal cancer and endometrial cancer, but not in head and neck cancer. Our results indicated that the RAD51 135G/C polymorphism was a candidate for susceptibility of cancer. The effect of the variants on the expression levels and the possible functional role of the variants in different cancers should be addressed in further studies.