Current evidence suggests that the aetiology of congenital gastrointestinal (GI) tract atresia is multifactorial, and not based solely on genetic factors. However, there are no established modifiable risk factors for congenital GI tract atresia. We used data from a Japanese nationwide birth cohort study launched in 2011, and examined whether fish consumption in early pregnancy was associated with congenital GI tract atresia. We analysed data of 89 495 women (mean age at delivery=31·2 years) who delivered singleton live births without chromosomal anomalies. Based on the results of the FFQ, we estimated the daily intake of fish and n-3 PUFA consumption in early pregnancy. We defined a composite outcome (oesophageal atresia, duodenal atresia, jejunoileal atresia and/or anorectal malformation) as congenital GI tract atresia. In this population, median fish intake was 31·9 g/d, and seventy-four cases of congenital GI tract atresia were identified. Fish consumption in early pregnancy was inversely associated with the composite outcome (multivariable-adjusted OR for the high v. low consumption category=0·5, 95 % CI 0·3, 1·0). For all the specific types of atresia, decreased OR were observed in the high consumption category, although not statistically significant. Reduced atresia occurrence was observed even beyond the US Food and Drug Administration’s recommended consumption of no more than 340 g/week. Also, n-3 PUFA-rich fish and n-3 PUFA consumptions tended to be inversely associated with atresia. Fish consumption in early pregnancy may be a preventive factor for congenital GI tract atresia.