To evaluate whether the accuracy and duration of registration for cone beam computed tomography (CBCT) and intraoral scans differ according to the method of registration and ratio of dental restorations to natural teeth.CBCT data and intraoral scans of eligible patients were grouped as follows according to the ratio of the number of dental restorations to the number of natural teeth (N): group 1, N = 0%; group 2, 0% < N < 50%; group 3, 50% ≤ N < 100%; and group 4, 100% ≤ N. Marker-free registration was performed with a deep learning-based platform and four implant planning software with different registration methods (two point-based, one surface-based, and one manual registration software) by a single operator, and the time consumption was recorded. Registration accuracy was evaluated by measuring the distances between the three-dimensional models of CBCT data and intraoral scans.A total of 36 patients, one jaw per patient, were enrolled. Although registration accuracy was similar, the time consumed for registration significantly differed for the different methods. The deep learning-based registration method consumed the least time. Greater proportions of dental restorations significantly reduced the registration accuracy for semi-automatic and deep learning-based methods and reduced the time consumed for semi-automatic registration.No superiority in registration accuracy was found. The proportion of dental restorations significantly affects the accuracy and duration of registration for CBCT data and intraoral scans.ClinicalTrials.gov Identifier: KCT0006710 CLINICAL RELEVANCE: Registration accuracy for virtual implant planning decreases when the proportion of dental restorations increases regardless of registration methods.