Background There is substantial debate as to whether moderate alcohol use during pregnancy could have subtle but important effects on offspring, by impairing later cognitive function and thus school performance. The authors aimed to investigate the unconfounded effect of moderately increased prenatal alcohol exposure on cognitive/educational performance.Methods We used mother-offspring pairs participating in the Avon Longitudinal Study of Parents and Children (ALSPAC) and performed both conventional observational analyses and Mendelian randomization using an ADH1B variant (rs1229984) associated with reduced alcohol consumption. Women of White European origin with genotype and self-reported prenatal alcohol consumption, whose offspring’s IQ score had been assessed in clinic (N = 4061 pairs) or Key Stage 2 (KS2) academic achievement score was available through linkage to the National Pupil Database (N = 6268), contributed to the analyses.Results Women reporting moderate drinking before and during early pregnancy were relatively affluent compared with women reporting lighter drinking, and their children had higher KS2 and IQ scores. In contrast, children whose mothers’ genotype predisposes to lower consumption or abstinence during early pregnancy had higher KS2 scores (mean difference +1.7, 95% confidence interval +0.4, +3.0) than children of mothers whose genotype predisposed to heavier drinking, after adjustment for population stratification.Conclusions Better offspring cognitive/educational outcomes observed in association with prenatal alcohol exposure presumably reflected residual confounding by factors associated with social position and maternal education. The unconfounded Mendelian randomization estimates suggest a small but potentially important detrimental effect of small increases in prenatal alcohol exposure, at least on educational outcomes.