Aim: A clinical study was done to assess the clinical diagnostic accuracy of cone beam computed tomography (CBCT) in detecting proximal cavitated carious lesions in order to determine whether cavitation should be reported when a CBCT examination is available. Materials and Methods: 79 adjacent proximal surfaces without restorations in permanent teeth were examined. Patients suspected to have carious lesions after a visual clinical and a bitewing examination participated in a CBCT examination (Kodak 9000 3D, 5 × 3.7 cm field of view, voxel size 0.07 mm). Ethical approval and informed consent were obtained according to the Helsinki Declaration. Radiographic assessment recording lesions with or without cavitation was performed by two observers in bitewings and CBCT sections. Orthodontic separators were placed interdentally between two lesion-suspected surfaces. The separator was removed after 3 days and the surfaces recorded as cavitated (yes/no), i.e. validated clinically. Differences between the two radiographic modalities (sensitivity, specificity and overall accuracy) were estimated by analyzing the binary data in a generalized linear model. Results: For both observers, sensitivity was significantly higher for CBCT than for bitewings (average difference 33%, p <0.001) while specificity was not significantly different between the methods (p = 0.19). The overall accuracy was also significantly higher for CBCT (p <0.001). Conclusion: CBCT was more accurate in detecting cavitation in proximal surfaces than bitewing radiographs; therefore a CBCT examination performed for other clinical applications should also be assessed for proximal surface cavities in teeth without restorations, and when detected, this pathology must be part of the dentist's report.