Purpose: To compare accuracy and complications of computer-assisted template-based implant placement using conventional impression and scan of a physical stone cast or intraoral scanning to rehabilitate partially edentulous patients. Materials and methods: Any partially edentulous patients with at least five residual teeth, requiring at least one implant to be planned on three-dimensional (3D) cone beam computed tomography (CBCT) scan according to a computer-assisted template-based protocol were enrolled. Patients were randomised according to a parallel-group design into two arms: intraoral digital impression (fully digital group) or conventional impression and scan model (conventional group). Implants were placed flapless or with a minimally invasive flap, and conventionally loaded after 5 months. Outcome measures were implant and prosthetic success, complications, accuracy and peri-implant marginal bone loss. Three deviation parameters (angular, horizontal and vertical) were defined to evaluate the discrepancy between the planned and placed implant positions. Results were compared using a mixed-model repeated-measures analysis of variance (α = 0.05). Results: Twenty patients (11 females and 9 males; mean age 44.6 years old) were randomised to the fully digital group (10 patients with 28 implants) or conventional group (10 patients with 29 implants). No patients dropped out. No implant or prosthesis failed up to 1 year after loading. One implant in the fully digital group was placed freehand due to limited inter-arch space. No biological or mechanical complications were experienced during follow-up. Difference between groups were not statistically significant (P = 0.999). The mean error in angle was 2.25 ± 1.41 degrees (range 0.30 to 5.00 degrees; 95% CI: 1.38 to 3.12 degrees) in the fully digital group and 2.10 ± 1.18 degrees (range 0.30 to 5.80 degrees; 95% CI: 1.37 to 2.83 degrees) in the conventional group. The difference was not statistically significant (−0.15 ± 1.63 degrees; range −3.20 to 2.90 degrees; 95% CI: −0.87 to 0.57 degrees; P = 0.668); in the horizontal plane (mesio-distal), the mean error was 0.52 ± 0.30 mm (range 0.10 to 1.10 mm; 95% CI: 0.33 to 0.70 mm) in the fully digital group and 0.44 ± 0.26 mm (range 0.10 to 0.90 mm; 95% CI: 0.27 to 0.60 mm) in the conventional group. The difference was not statistically significant (−0.08 ± 0.38 degrees; range −1.0 to 0.60 degrees; 95% CI: −0.32 to 0.16 degrees; P = 0.279); in the vertical plane (apico-coronal), the mean error was 0.58 ± 0.44 mm (range 0.00 to 1.60 mm; 95% CI: 0.31 to 0.85) in the fully digital group and 0.46 ± 0.34 mm (range 0.00 to 1.20 mm; 95% CI: 0.25 to 0.67) in the conventional group. The difference was not statistically significant (−0.12 ± 0.59 degrees; range −1.20 to 1.00 degrees; 95% CI: −0.49 to 0.24 degrees; P = 0.250). One year after loading, the mean marginal bone loss was 0.14 ± 0.12 mm (range −0.10 to 0.40 mm; 95% CI: 0.07 to 0.21 mm) in the fully digital group and 0.18 ± 0.13 mm (range −0.10 to 0.60 mm; 95% CI: 0.09 to 0.26 mm). The difference was not statistically significant (−0.04 ± 0.19 mm; range -0.50 to 0.30 mm; 95% CI: −0.16 to 0.08 mm; P = 0.294). Conclusions: With the limitations of the present trial, implant rehabilitations planned using intraoral digital impressions showed similar results compared to conventional impression and scan model. Digital impression may be a viable option for the rehabilitation of partial edentulous patients when computer-guided template-assisted implant placement is used.