Purpose: To compare the three-dimensional (3D) accuracy of conventional direct implant impressions with digital implant impressions from three intraoral scanners, as well as different implant levels-bone level (BL) and tissue level (TL). Materials and Methods: Two-implant master models were used to simulate a threeunit implant-supported fixed dental prosthesis. Conventional test models were made with direct impression copings and polyether impressions. Scan bodies were hand-tightened onto master models and scanned with the three scanners. This was done for the TL and BL test groups, for a total of eight test groups (n = 5 each). A coordinate measuring machine measured linear distortions (dx, dy, dz), global linear distortion (dR), angular distortions (dϴy, dϴx), and absolute angular distortions (Absdϴy, Absdϴx) between the master models, test models, and .stl files of the digital scans. Results: The mean dR ranged from 35 to 66 μm; mean dϴy angular distortions ranged from -0.186 to 0.315 degrees; and mean dϴx angular distortions ranged from -0.206 to 0.164 degrees. Two-way analysis of variance showed that the impression type had a significant effect on dx, dz, and Absdϴy, and the implant level had a significant effect on dx and Absdϴx (P .05). Among the BL groups, the mean dR of the conventional group was lower than and significantly different from the digital test groups (P = .010), while among the TL groups, there was no statistically significant difference (P = .572). Conclusion: The 3D accuracy of implant impressions varied according to the impression technique and implant level. For BL test groups, the conventional impression group had significantly lower distortion than the digital impression groups. Among the digital test groups, the TR system had comparable mean linear and absolute angular distortions to the other two systems but exhibited the smallest standard deviations. Keywords: accuracy, coordinate measuring machine, digital, implant impression, three-dimensional, 3D