Stroke represents a major burden in patients with type 2 diabetes, yet this cerebrovascular complication has been less carefully investigated than the risk of cardiovascular mortality, heart failure and renal disease. Some data suggested that glucagon-like peptide-1 receptor agonists (GLP-1RAs) exert a better protection against stroke than sodium-glucose cotransporter 2 inhibitors (SGLT2is). However, this conclusion was derived from indirect comparisons in absence of any head-to-head randomised controlled trial (RCT). The present comprehensive review compares the effects of SGLT2is versus GLP-1RAs on nonfatal and fatal/nonfatal strokes in network meta-analyses of RCTs (mostly cardiovascular outcome trials) versus placebo, on the one hand, and in real-life observational cohort studies, on the other hand. Whereas network meta-analyses of placebo-controlled RCTs confirm a slight but significant (in 11 out of 13 meta-analyses) higher incidence of stroke in patients treated with SGLT2is compared with those treated with GLP-1RAs, a large majority of retrospective observational cohort studies (19 out of 21) failed to find any significant difference in the risk of stroke between the two pharmacological classes. Available, yet limited, findings suggest that SGLT2is may be more efficacious against haemorrhagic than ischaemic strokes, in patients at risk for atrial fibrillation and in patients with chronic kidney disease.