Results of studies investigating the long-term effects of intra-uterine exposure to antiepileptic drugs (AEDs) on cognitive functioning are limited and conflicting. To estimate intellectual development of children prenatally exposed or unexposed to AEDs by assessing IQ scores in a systematic review and meta-analysis. A literature search using Pubmed, EMBASE and Google Scholar from inception to 30 April 2009 was performed to identify all original cohort studies that investigated cognitive functioning after in utero exposure to AEDs. Studies had to include at least one group exposed to an AED and one unexposed group. Data from drug exposed and unexposed controls were combined using a random effects model. Eleven studies met the inclusion criteria. Eight studies (three for valproic acid and five for carbamazepine) evaluated IQ as a measure of cognitive development. IQ was assessed by the Wechsler, Bayley or McCarthy intelligence scales, depending on age. One study investigated phenytoin and one study investigated phenobarbital (phenobarbitone). Because one study was reported in two different publications, seven studies were included in the meta-analysis. In total, the seven selected studies included 67 children exposed in utero to valproic acid and 151 exposed to carbamazepine, and 494 unexposed controls born to healthy women or to women with untreated epilepsy. The mean full-scale IQ (FSIQ), verbal IQ (VIQ) and performance IQ (PIQ) scores in children exposed to valproic acid in utero were 83.9 (95% CI 64.2, 103.6), 93.7 (95% CI 72.6, 114.7) and 88.3 (95% CI 69.9, 106.9), respectively. The mean FSIQ, VIQ and PIQ scores in the control group were 102 (95% CI 90, 116), 101 (95% CI 87, 114) and 99 (95% CI 90, 117), respectively. The mean FSIQ, VIQ and PIQ were all significantly lower in the valproic acid group compared with the unexposed group. The FSIQ and VIQ of children exposed to carbamazepine were not statistically different from those of the unexposed control group. In a sub-analysis of carbamazepine exposure in three studies using the Wechsler intelligence scale, PIQ was significantly lower in children exposed to carbamazepine than in unexposed children. Although our analysis revealed no evidence that untreated maternal epilepsy was associated with a lower IQ in the child, there may have been confounding factors, such as milder epilepsy, in this group. Exposure to valproic acid in pregnancy is associated with significantly reduced intelligence in children whose mothers were treated for epilepsy. Exposure to carbamazepine in pregnancy does not appear to be associated with reduced FSIQ and VIQ in children, although PIQ was lower in the sub-analysis. Clinicians should inform families of the potential cognitive adverse effects of valproic acid. More studies are needed to corroborate these findings.