Abstract Objective The performance of existing diagnostic criteria for Behçet’s disease (BD) is usually evaluated by comparison with expert opinions, which may be limited by misclassification and disagreement among experts. We aim to evaluate these criteria in the absence of a gold standard. Methods We obtained two datasets involving possible BD and other mimickers from a uveitis registry using case–cohort and nested case–control analyses, respectively. With a Bayesian inference approach, the sensitivity and specificity of International Study Group (ISG) and International Criteria for Behçet’s Disease (ICBD) criteria were simultaneously estimated when true BD state was unknown. Results A total of 2440 and 2224 participants were included in case–cohort and nested case–control analyses, respectively. In case–cohort analysis, with scores of ≥4 for BD diagnosis, ICBD criteria showed higher sensitivity (median 97.6%; 95% credible interval 96.9, 98.2) than ISG criteria (median 90.0%; 95% credible interval 88.8, 91.2) but had lower specificity (median 90.8%; 95% credible interval 89.4, 92.1) than ISG criteria (median 98.8%; 95% credible interval 98.3, 99.3). With scores of ≥5 for diagnosis, ICBD criteria demonstrated higher sensitivity (median 97.5%; 95% credible interval 96.8, 98.1) and specificity (median 99.6%; 95% credible interval 99.3, 99.8) than the sensitivity (median 92.3%; 95% credible interval 91.2, 93.3) and specificity (median 98.8%; 95% credible interval 98.2, 99.2) for ISG criteria. The highest diagnostic consistency was observed between ISG criteria and ICBD criteria with scores of ≥5 for diagnosis (Kappa = 0.999; P < 0.001). Nested case–control analysis showed similar results. Conclusion ICBD criteria showed optimum discriminatory properties in sensitivity and specificity with scores of ≥5 for BD diagnosis in uveitis. The diagnostic threshold of ICBD criteria could be considered adjustable according to medical specialty, disease prevalence and local practice characteristics.