The association of two cytochrome P4501A1 (CYP1A1) polymorphisms, m1 (T6235C transition) and m2 (A4889G transition), with gastric cancer risk is inconclusive. We conducted a meta-analysis of all available studies to evaluate the potential role of the polymorphisms and their interactions with tobacco smoking in gastric cancer susceptibility. Published literature from PubMed was retrieved by two investigators independently. Fourteen case-control studies with 2,032 gastric cancer cases and 5,099 controls were selected. A fixed effects model or a random-effects model was used to estimate the odds ratio (OR) for the CYP1A1 polymorphisms and the occurrence of gastric cancer. Significant associations between CYP1A1 m1 and m2 polymorphisms and gastric cancer susceptibility were not observed in all genetic models in the overall analyses. Subgroup analyses by ethnicity and source of controls did not reveal significant associations with gastric cancer risk. Stratification analysis by smoking status found that carriers of the heterozygous and homozygous m1 genotypes decreased the susceptibility of gastric cancer among ever-smokers (pooled OR = 0.56, 95 % CI 0.36-0.89, fixed effects). In contrast, the m2 genotypes (G/G and A/G) did not show any relevance to gastric cancer risk among the smoking population (pooled OR = 1.30, 95 % CI 0.84-2.00, fixed effects). Overall, we found that the CYP1A1 polymorphism itself, either m1 or m2, did not represent an independent genetic risk factor influencing gastric cancer. However, subgroup analyses suggest that carriers of the heterozygous and homozygous m1 genotype who are exposed to tobacco smoke have a significantly lower risk of developing gastric cancer. To explain the observed reduction of gastric cancer risk, we proposed a novel hypothesis of "observation bias". This hypothesis is also applicable to explain the combined effects of other genetic polymorphisms and environmental factors on the risk of developing cancers, and the rationality of the hypothesis needs to be further investigated.