Abstract Objectives The basic prerequisite for the production of dental restorations by means of CAD/CAM technologies is the data acquisition (digitization). Currently, two methods are available, i.e. the extraoral digitization of master casts and the direct intraoral data acquisition. However, it seems to be beneficial to immediately digitize impressions directly at the dental office in order to combine the high precision of mechanical digitizing methods and to shorten the production process. The aim of this study was to investigate the measurement uncertainty (±2 σ) and the three-dimensional accuracy of the immediate tactile in-office digitization of dental impressions and of the mechanical digitizing of ceramic master dies using a high-precision touch-probe digitizer. Methods The experimental set-up consisted of ceramic master dies representing tooth 13 and 36 as well as their identical virtual models (CAD models). Fifteen one-step putty-wash impressions were taken from each tooth. The impressions as well as the ceramic master dies were digitized applying a standardized procedure. The datasets were aligned to the corresponding CAD models; then, a computer-aided three-dimensional analysis was performed. Results The digitizing of the dental impressions showed a measurement uncertainty of 5.8, mean positive deviations between 27 and 28 μm, and mean negative deviations between −21 and −31 μm. The digitizing of the ceramic master dies showed a measurement uncertainty of 2.8, mean positive deviations between 7.7 and 9.1 μm, and mean negative deviations between −8.5 and −8.8 μm. Conclusion Mechanical digitizers show a very low measurement uncertainty and a high precision. The immediate tactile in-office digitization of impressions cannot be recommended as adequate data acquisition method for CAD/CAM restorations. It is recommendable to digitize clinical sites extraorally, i.e. after taking an impression and fabricating a model cast thereof.