ABSTRACT Background The diagnostic criteria for severe ischemic stroke (SIS) vary across studies. There are no generally accepted diagnostic criteria for SIS and no consensus on which criteria are most appropriate in clinical practice. We aimed to evaluate the clinical significance of these criteria by comparing their associations with death/disability in an independent clinical dataset. Methods This was a retrospective cohort study that consecutively included ischemic stroke patients admitted within 7 days of onset. We conducted a literature review to determine the commonly used diagnostic criteria for SIS. We used the sensitivity, specificity, and Youden index to assess the relationship of the criteria with the poor outcomes and calculated the area under the receiver‐operator characteristics curve (AUC). Results We identified 10 commonly used diagnostic criteria for SIS through literature review, including four clinical criteria, three imaging criteria, and three composite criteria. A NIHSS score ≥ 15 had the maximal Youden index (0.395) and AUC (0.697, 95% CI 0.641–0.754) for predicting poor outcomes at 3 months, followed by NIHSS 1a ≥ 1 (Youden index 0.387; AUC 0.693, 95% CI 0.637–0.750). Despite the poor comprehensive diagnostic ability, the composite criteria had a high specificity of over 97% and a negative predictive value of nearly 84%. For death or transfer to the intensive care unit during hospitalization, NIHSS 1a ≥ 1 had the maximal Youden index (0.596) and AUC (0.798, 95% CI 0.728–0.869), followed by NIHSS ≥15 (Youden index 0.539; AUC 0.769, 95% CI 0.692–0.847). Conclusions The severity of stroke assessed by NIHSS ≥ 15 and the impaired consciousness defined by NIHSS 1a ≥ 1 showed the best diagnostic ability for SIS. Our results provide scientific evidence to identify a simple and unified diagnostic criterion for SIS, which will help to enable cross‐comparison between research and facilitate communication of case series in clinical practice.