Many studies have been carried out to confirm the relationship between androgen receptor gene CAG repeat polymorphism and polycystic ovary syndrome (PCOS), without consistent results. Hence we conducted the current study to research this relationship.224 Chinese Han women with PCOS and 223 in vitro fertilization and embryo transplantation (IVF-ET) infertile women with tubal factor or male infertility served as the controls were recruited in our study. PCR-based assays were applied to genotype the (CAG)n repeat alleles. A meta-analysis including 1,536 PCOS patients and 1,807 controls was conducted to produce a pooled estimate.We observed that the CAG bi-allelic mean lengths were similar in PCOS patients and controls (22.65 ± 2.5 vs. 23.09 ± 2.1, P = 0.116). When CAG bi-allelic were divided into two categories (mean repeats ≤22,22), the short AR-CAG bi-allelic showed more frequent in PCOS group than in controls (56.25% vs 29.14%, P 0.001). Further analysis presented that, in PCOS, there was a lower mean CAG repeat lengths in mean bi-allelic lengths (22.3 ± 2.5 vs. 23.9 ± 2.2, P = 0.008) and long bi-allelic lengths (24.3 ± 1.4 vs. 25.9 ± 1.6, P = 0.05) among patients with testosterone less than 0.7 ng/ml compared with those whose testosterone was more than 0.7 ng/ml. Besides, the testosterone were positively correlated with the CAG polymorphism (r = 0.237, P = 0.008), which accorded with our meta-analysis results.The distribution of AR-CAG allele differed between PCOS patients and controls, and polymorphism of CAG repeat lengths may contribute to hyperandrogenism in PCOS.