Background Endovascular embolization is an accepted treatment modality for brain arteriovenous malformations (bAVM); however, treatment outcomes are highly variable, warranting accurate prediction for adequate patient selection. Several predictive scores have been proposed for this purpose. The objective of this study was to externally validate these scores for embolization of bAVM.Methods This study involved bAVM patients treated with transarterial embolization. Endovascular predictive scores were identified through literature search. Relevant data for scoring of included patients was extracted. Primary study outcomes were radiological cure and neurological complications. The performance of the scores was evaluated by analyzing calibration (z-scores from logistic regression), discrimination (area under the receiver operating characteristic curve, AUROC), and classification (Youden's index and corresponding sensitivity and specificity). Additionally, sensitivity analyses were performed restricting the study population by size, location, and embolization intent.Results A total of 198 bAVM (190 patients) were included. The rates of radiological cure and neurological complications were 18.2% and 14.1%, respectively. The literature search identified seven predictive scores. In the overall analysis, the Toronto score showed the best performance for radiological cure (AUROC 0.905). No significant difference was observed between the performance of the assessed scores for neurological complications. The sensitivity analysis showed improved performance of most scores. The Toronto score exhibited the highest performance for radiological cure (AUROC 0.857). The AVM Embolization Prognostic Risk Score (AVMEPRS) showed the highest performance for neurological complications (AUROC 0.751). The AVM Embocure Score (AVMES) showed fair to good performance for both efficacy and safety outcomes.Conclusion Among the selected scores, the Toronto, AVMEPRS, and AVMES scores showed the best performances.