Aim To assess observer performance in detecting endodontic complications using three different cone beam computed tomography (CBCT) units with and without the application of artefact reduction modes. Methodology The study involved 40 freshly extracted human mandibular teeth (n = 10 per group) and divided randomly into four endodontic complication groups. Group 1) Instrument fracture; Group 2) Strip perforation; Group 3) Canal underfilling; and Group 4) Canal overfilling. Images of each tooth were obtained using three different CBCT units offering artefact reduction algorithms: the ProMax 3D Max, the Pax Flex 3D and the Dentri S. Four observers evaluated the images for the presence/absence of the four simulated endodontic complications. Weighted kappa coefficients and intra-class correlation coefficients (ICCs) were calculated to reveal the intra- and inter-observer agreement for each imaging mode, respectively. Receiver operating characteristic (ROC) analysis was used to evaluate the observers' performance. DeLong tests were used to compare the results for each image mode and observer using a significance level of alpha = 0.05. Results In each of the four simulated endodontic complication groups, no significant differences were observed with and without application of artefact reduction for any of the three CBCT units tested. Only two significant differences were detected, and both were between the ProMax 3D Max at low mode AR and ProMax 3D Max without AR: observer 2 in group 1 (P = 0.0001) and observer 4 in group 4 (P = 0.0256). Conclusion For each of the three CBCT units tested, application of artefact reduction for detecting endodontic complications is not recommended as a routine tool.