Because of its ability to capture hard structures behind soft tissue, ultrasound-based micro-scanning may be a promising alternative for taking digital impressions of teeth, especially in the case of subgingival margin preparations. The aim of this study was to assess the accuracy of ultrasound impressions taken of subgingivally prepared teeth compared with digital optical impressions. Ten extracted human teeth (7 pre-molars, 3 molars) were prepared for crowns with chamfer finish line and then digitized using two different intra-oral scanners (Cara Trios, 3 Shape, Heraeus Kulzer, Hanau, Germany; and Lava COS; 3M ESPE, Seefeld, Germany) and one extra-oral scanner (Cares CS2, Straumann, Basel, Switzerland). Afterward, the preparation margin was covered with porcine gingiva (thickness ranged between 0.3 and 0.9 mm), and every sample was scanned with a high-frequency ultrasound scanner under experimental subgingival conditions. Optical scanning processes were performed without gingiva. The data sets were superimposed on each other for pairwise comparisons, and deviations between different scans were determined using a 3-D evaluation software (CloudCompare). Kruskal-Wallis and post hoc tests (Dunn-Bonferroni) were applied to detect significant differences at p ≤ 0.05. The ultrasound scanner was able to detect subgingival preparation margins. Mean deviations for all comparisons ranged from 12.34 to 46.38 µm. There were no statistically significant differences between superimpositions of intra-oral and extra-oral scans (Trios-Lava, Lava-CS2, Trios-CS2), whereas in comparisons between intra-/extra-oral scans and ultrasound scans, mean deviations were statistically significantly higher. There were no significant differences with respect to type of tooth (pre-molar and molar). However, gingiva thickness was significantly correlated with the quality of the ultrasound scan; thin layers had better image quality than thicker layers. Ultrasound was able to scan tooth preparation margins covered with gingiva, although with less accuracy than achieved by conventional optical scanners (non-covered margins). Gingiva thickness may play an important role in ultrasound scan accuracy.