We aimed to perform a systematic review including a meta-analysis to evaluate the overall accuracy of visual methods for detecting carious lesions and to identify possible sources of heterogeneity among the studies included. Two reviewers searched PubMed, Embase, Scopus, and other sources through July 2014 to identify published and nonpublished studies in English. Studies of visual inspection were included that 1) assessed accuracy of the method in detecting caries lesions; 2) were performed on occlusal, proximal, or free smooth surfaces in primary or permanent teeth; 3) had a reference standard; and 4) reported sufficient data about sample size and accuracy of methods. The data were used to calculate the pooled sensitivity, specificity, diagnostic odds ratio, and summary receiver operating characteristics curve. Heterogeneity of the studies was also assessed. A total of 102 manuscripts (from 5,808 articles initially identified) and 1 abstract (from 168) met the inclusion criteria. In general, the analysis demonstrated that the visual method had good accuracy for detecting caries lesions. Although laboratory and clinical studies have presented similar accuracy, clinically obtained specificity was higher. We also observed moderate to high heterogeneity and evidence of publication bias in most papers. Moreover, studies employing widely recognized visual scoring systems presented significantly better accuracy as compared to studies that used their own criteria. In conclusion, visual caries detection method has good overall performance. Furthermore, although the identified studies had high heterogeneity and risk of bias, the use of detailed and validated indices seems to improve the accuracy of the method.