AimsThe objective of this umbrella review and meta‐analysis was to evaluate the effect of diabetes on risk of dementia, as well as the mitigating effect of antidiabetic treatments.Materials and MethodsWe conducted a systematic umbrella review on diabetes and its treatment, and a meta‐analysis focusing on treatment. We searched MEDLINE/PubMed, Embase, PsycINFO, CINAHL and the Cochrane Library for systematic reviews and meta‐analyses assessing the risk of cognitive decline/dementia in individuals with diabetes until 2 July 2023. We conducted random‐effects meta‐analyses to obtain risk ratios and 95% confidence intervals estimating the association of metformin, thiazolidinediones, pioglitazone, dipeptidyl peptidase‐4 inhibitors, α‐glucosidase inhibitors, meglitinides, insulin, sulphonylureas, glucagon‐like peptide‐1 receptor agonists (GLP1RAs) and sodium‐glucose cotransporter‐2 inhibitors (SGLT2is) with risk of dementia from cohort/case‐control studies. The subgroups analysed included country and world region. Risk of bias was assessed with the AMSTAR tool and Newcastle‐Ottawa Scale.ResultsWe included 100 reviews and 27 cohort/case‐control studies (N = 3 046 661). Metformin, thiazolidinediones, pioglitazone, GLP1RAs and SGLT2is were associated with significant reduction in risk of dementia. When studies examining metformin were divided by country, the only significant effect was for the United States. Moreover, the effect of metformin was significant in Western but not Eastern populations. No significant effect was observed for dipeptidyl peptidase‐4 inhibitors, α‐glucosidase inhibitors, or insulin, while meglitinides and sulphonylureas were associated with increased risk.ConclusionsMetformin, thiazolidinediones, pioglitazone, GLP1RAs and SGLT2is were associated with reduced risk of dementia. More longitudinal studies aimed at determining their relative benefit in different populations should be conducted.