A large odontometric reference database is essential for a clear and reliable differentiation between the teeth of dogs and wolves found in archaeological contexts. However, the data compilation of skeletal material is a slow process because access to dog remains may be difficult. The expansion of a dental database could benefit from computed tomography (CT) scans performed on live dogs during medical appointments and archived in veterinary clinics. To enable it, however, the reliability and accuracy of digital measurements must first be validated. This is the main objective of this paper. We tested the accuracy of dental measurements obtained from CT scans of deceased modern dogs and one Mesolithic dog. CT measurements were obtained using multi-planar reconstruction (MPR) (bi-dimensional images) and three-dimensional volume rendering (3D VR), either using bone or soft tissue windows. Then, measurements were compared with data obtained with a conventional caliper (Fischer Darex®, France). The sample comprised 25 maxillary fourth premolars and 17 maxillary canines. Measurements included the mesiodistal length (MDL) and the vestibular palatine width (VPW). Intra- and inter-observer variations were assessed by calculating the technical error of measurement (TEM). This approach was also used to assess the level of agreement between both techniques, complemented by a Bland-Altman analysis. Intra-observer (TEM = 0.52% to 5.98%) and inter-observer variations (TEM = 3.15% to 7.14%) were relatively small for most standard measurements, windows, and CT approaches. In the case of the MPR approach, the soft tissue window provided less observer variation on the fourth premolar; the same occurred for the bone window regarding the canine. The soft tissue window generally provided less observer variation with the VR approach. CT and caliper agreement ranged from 2.89% to 7.31% (MPR), and from 2.92% to 9.42% (3D VR). At least 91% of the CT measurements were within the Bland-Altman 95% confidence interval. We were able to carry out a thorough odontometric study of the Muge dog skull – a specimen dated to the Mesolithic period, using the MPR approach (with both bone and soft tissue windows). Both the MPR and VR CT scan approaches were validated, thus confirming digital data as an alternative way to generate more comprehensive odontometric references.