Introduction Several studies have reported on the maternal age‐associated risks of congenital anomalies. However, there is a paucity of studies with comprehensive review of anomalies. We aimed to quantify the risk of birth defects in children born to middle‐aged mothers compared with that in children born to young or older mothers. Material and methods We classified maternal ages into three groups: young (<20 years old), middle (20–34 years old) and older age (≥35 years old). Observational studies that met our age criteria were eligible for inclusion. The articles searched using the Embase and MEDLINE databases were those published from 1989 to January 21, 2021. The Newcastle–Ottawa scale was used to assess the risk of bias. If heterogeneity exceeded 50%, the random effect method was used; otherwise, the fixed‐effect method was used. Prospero registration number: CRD42021235229. Results We included 15 cohort, 14 case–control and 36 cross‐sectional studies. The pooled unadjusted odds ratio (95% CI) of any congenital anomaly was 1.64 (1.40–1.92) and 1.05 (0.95–1.15) in the older and young age groups, respectively (very low quality of evidence). The pooled unadjusted odds ratio of chromosomal anomaly was 5.64 (5.13–6.20) and 0.69 (0.54–0.88) in the older and young age groups, respectively. The pooled unadjusted odds ratio of non‐chromosomal anomaly was 1.09 (1.01–1.17) and 1.10 (1.01–1.21) in the older and young age groups, respectively (very low quality of evidence). The incidence of abdominal wall defects was increased in children of women in the young maternal age group. Conclusions We identified that very low quality evidence suggests that women in the older maternal age group had increased odds of having children with congenital anomalies compared with those in the 20–34 year age group. There was no increase in odds of children with congenital anomalies in women of <20 year age group except for abdominal defects compared with those in the 20–34 year age group. The results stem from very low quality evidence with no adjustment of confounders.