IntroductionReview of videos (without electroencephalography) to differentiate epileptic seizures (ES) from non-epileptic spells (NES) may be helpful where epilepsy monitoring is not feasible. Previous studies of video-based diagnosis have suffered from variable accuracy, sensitivity, and specificity. MethodsWe systematically reviewed relevant literature in PubMed, Embase, and Web of Science from inception to September 2022, identifying articles that reported on the video-based diagnosis of ES and NES. In primary analysis, for each study, the most expert group was chosen when different groups of reviewers classified the videos (e.g., epilepsy specialists and general neurologists). In secondary analysis, we compared the diagnostic accuracy of different expertise levels (e.g., epileptologists, general neurologists, residents, medical students). Meta-analysis was performed to obtain pooled estimates of reliability measures. ResultsFrom 5245 articles identified, 13 met the inclusion criteria, with cumulative data from 683 patients (696 videos) reviewed by 95 independent reviewers in primary analysis. Video alone had a strong ability to differentiate ES from NES as evidenced by the following metrics- area under the curve- 0.9 (considered “outstanding”), sensitivity- 82.2% (95% Confidence Interval [C.I], 80.2%-84.0%), specificity- 84.7% (C.I., 82.8%-86.5%), and diagnostic odds ratio- 24.7 (C.I., 11.5–52.9). The secondary analysis showed reviewer-dependent accuracy with epileptologists showing the highest accuracy (DOR 81.2, C.I., 90.0%-94.6%). ConclusionsVideo alone has reliable diagnostic performance for differentiating ES from NES. Meta-analysis limitations include inter-study heterogeneity including variable video quality and reviewer expertise. Combined video-EEG remains the gold standard for the diagnosis of epilepsy and NES.