Purpose: To compare outcomes for immediately loaded one-piece implants (OPI), placed flapless, and conventionally loaded two-piece implants (TPI), placed after two-stage flapped surgery. Materials and methods: Thirty-eight participants were randomised to receive either one OPI (19 participants, OPI group) or one TPI (19 participants, TPI group) inserted in the posterior mandible with a torque of at least 35 Ncm according to a parallel group design. OPI were immediately loaded with non-occluding temporary crowns. After 3 months, TPI were exposed and implants in both groups were occlusally loaded with zirconia crowns. Outcome measures were implant failure, prosthesis failure, any complication and changes of probing pocket depth (PPD), plaque index (PI), gingiva index (GI), and peri-implant marginal bone level, recorded by unblinded assessors. Results: Three years after occlusal loading, three participants dropped out from the TPI group. There were no statistically significant differences between the groups with regard to participants with implant failure (OPI group 1/19; TPI group 0/16; difference in proportions (DIP), ?5.3%; 95% confidence interval (CI) ?15.3 to 4.8; P = 1.000), prosthesis failure (OPI group 3/19, TPI group 5/16; DIP, 15.5%; 95% CI ?12.6 to 43.5; P = 0.424), any complication (OPI group 6/19, TPI group 5/16; DIP, ?0.3%; 95% CI ?31.2 to 30.5; P = 1.000), or changes of PPD (P =0.174), PI (P = 0.222), or GI (P = 0.415). Veneer chipping accounted for most prosthesis failures and complications. On average OPI and TPI lost 1.34 mm and 0.67 mm of marginal bone, respectively, the difference between groups being statistically significant (mean difference, 0.66 mm; 95% CI ?0.02 to 1.34; P = 0.024) in favour of TPI implants. Conclusions: Both implant procedures might be viable in the short term, but statistically significantly more bone loss might be indicative of future problems with OPI. Because of the high incidence of chipping, veneered zirconia crowns cannot be recommended on posterior implants.