This is a delayed-type cross-sectional prospective accuracy study nested in a randomized clinical trial. The aim was to investigate the diagnostic accuracy of two visual criteria for caries lesions detection around restorations in primary teeth: the International Dental Federation (FDI) criteria, considering adaptation, staining, and the presence of caries, and the Caries Associated with Restorations and Sealants (CARS) system. For this, one examiner made the diagnosis and subsequent treatment decision using visual assessment in 163 children (3-10 years old) with both FDI and CARS criteria. The order of criteria used was defined by randomization. The reference standard was composed of two approaches: (1) the presence of carious tissue after restoration removal and (2) the presence of caries lesions after 6 and 12 months of follow-up. Sensitivity, specificity, and accuracy parameters were calculated at the dentin threshold. Poisson multilevel regression analyses were performed to evaluate the association of the diagnostic methods and other explanatory variables with the outcomes. Of the 651 restorations included, 480 were evaluated by the reference standard methods and were analyzed. The CARS system presented higher accuracy (0.721) than those obtained with FDI recurrence of caries (0.702), FDI marginal adaptation (0.700), and FDI marginal staining criteria (0.681). The FDI marginal staining showed the study's lowest sensitivity (0.280) and accuracy (0.681) values. The specificity values of FDI recurrence of caries and FDI marginal adaptation were lower than the CARS system. Restorations assessed after the follow-up period resulted in lower sensitivity but higher specificity than those replaced after initial evaluation. In conclusion, the CARS system is more accurate in detecting caries around restorations in primary teeth than the FDI system, in general. However, the FDI recurrence of caries and FDI marginal adaptation present similar performance to the CARS system when the dentin threshold is considered. On the other hand, marginal staining is not an accurate parameter to evaluate caries around restorations.