The aim of this in vivo study was to evaluate the performance of fluorescence-based devices in detecting occlusal caries lesions in primary molars compared with conventional methods. Two examiners assessed 44 occlusal surfaces of first and second primary molars in 20 patients using two fluorescence devices: DIAGNOdent (LF) and DIAGNOdent pen (LFpen). Teeth were also assessed by visual examination and bitewing radiograph. Histological examination served as the gold standard after extraction. By using the McNemar test, the sensitivity, specificity, accuracy, and area under the receiver operating curve were calculated as outer enamel (D1), inner enamel (D2) and dentine caries (D3) lesion thresholds. The intra- and inter-examiner reproducibility were calculated using the Cohen's unweighted kappa statistics. At the D1 threshold, the LFpen sensitivity was statistically higher than LF and radiographic examination (P<0.001), whereas there was no statistically significant difference among the groups at the D2 and D3 thresholds (0.05). All methods demonstrated the highest sensitivity values at D3. At the D1 and D2 thresholds, there were no significant differences between the LFpen specificity and the other methods. All methods presented similar performance in detecting all lesions considering the area under the receiver operating curve. The LFpen showed better performance than LF. Furthermore, visual examination and the LFpen device seem to be sufficient for detection of occlusal caries in primary molars.