There is mixed evidence on the link between autism spectrum disorder (ASD) and diabetes. We conducted the first systematic review/meta-analysis on their association. Based on a pre-registered protocol (PROSPERO: CRD42021261114), we searched Pubmed, Ovid, and Web of Science databases up to 6 December 2021, with no language/type of document restrictions. We assessed study quality using the Newcastle-Ottawa Scale (NOS). We included 24 studies (total: 3427,773 individuals; 237,529 with ASD and 92,832 with diabetes) in the systematic review and 20 in the meta-analysis (mean stars number on the NOS: 5.89/10). There was a significant association, albeit characterized by significant heterogeneity, when pooling unadjusted OR (1.535, 95% CI = 1.109-2.126), which remained significant when restricting the analysis to children and type 2 diabetes, but became non-significant when considering adjusted ORs (OR: 1.528, 95% CI = 0.954-2.448). No significant prospective association was found (n = 2) on diabetes predicting ASD (HR: 1.232, 0.826-11.837). Therefore, the association between ASD and diabetes is likely confounded by demographic and clinical factors that should be systematically investigated in future studies.