Cone beam computed tomography (CBCT) has often been used to determine the quality of craniofacial bone structures through the determination of mineral density, which is based on gray scales of the images obtained. However, there is no consensus regarding the accuracy of the determination of the gray scales in these exams. This study aims to provide a literature review concerning the reliability of CBCT to determine bone mineral density. The gray values obtained with CBCT show a linear relationship with the attenuation coefficients of the materials, Hounsfield Units values obtained with medical computed tomography, and density values from dual energy X-ray absorciometry. However, errors are expected when CBCT images are used to define the quality of the scanned structures because these images show inconsistencies and arbitrariness in the gray values, particularly when related to abrupt change in the density of the object, X-ray beam hardening effect, scattered radiation, projection data discontinuity-related effect, differences between CBCT devices, changes in the volume of the field of view (FOV), and changes in the relationships of size and position between the FOV and the object evaluated. A few methods of mathematical correction of the gray scales in CBCT have been proposed; however, they do not generate consistent values that are independent of the devices and their configurations or of the scanned objects. Thus, CBCT should not be considered the examination of choice for the determination of bone and soft tissue mineral density at the current stage, particularly when values obtained are to be compared to predetermined standard values. Comparisons between symmetrically positioned structures inside the FOV and in relation to the exomass of the object, as it occurs with the right and left sides of the skull, seem to be viable because the effects on the gray scale in the regions of interest are the same.