Background and purposeThe DAWN trial recently showed compelling evidence in treating late window and wake‐up stroke patients with thrombectomy using a clinical‐imaging mismatch. The aim was to evaluate the results of thrombectomy for unknown‐onset strokes (UOS) treated in our centres after a diffusion weighted imaging/fluid attenuated inversion recovery (DWI‐FLAIR) mismatch based selection.MethodsA multicentre, cohort study was performed of consecutive UOS treated by thrombectomy between 2012 and 2016. UOS with proximal anterior circulation occlusion discovered beyond 6 h from ‘last seen normal’ were compared with known‐onset strokes (KOS) for whom thrombectomy was started within 6 h from onset. Time intervals were recorded from first time found abnormal. Results were adjusted for age, diabetes, hypertension, National Institutes of Health Stroke Scale, site of occlusion, DWI Alberta Stroke Programme Early CT Score, intravenous thrombolysis and use of general anaesthesia.ResultsAmongst 1246 strokes with anterior circulation occlusion treated by thrombectomy, 277 were UOS, with a ‘last time seen well’ beyond 6 h and DWI‐FLAIR mismatch, and 865 were KOS who underwent groin puncture within 6 h. Favourable outcome was achieved less often in UOS than KOS patients (45.2% vs. 53.9%, P = 0.022). After pre‐specified adjustment, this difference was not significant (adjusted relative risk 0.91; 95% confidence interval 0.80–1.04; P = 0.17). No differences were found in secondary outcomes. Time intervals from first found abnormal were significantly longer in UOS.ConclusionThrombectomy of UOS with anterior circulation occlusion and DWI‐FLAIR mismatch appears to be as safe and efficient as thrombectomy of KOS within 6 h from onset. This pattern of imaging could be used for patient selection when time of onset is unknown.