Virtual attachment removal (VAR) is a process used in digital dentistry to prefabricate retainers by removing attachments from a digital scan before the removal of clear aligner attachments. This study aims to evaluate the accuracy of VAR in the maxillary arch. 110 teeth were analyzed from a sample of 54 maxillary scans from 25 subjects. Models with attachments were virtually debonded using Meshmixer® and superimposed over the control group in MeshLab. Vector Analysis Module was used to calculate 3D distances on the buccal surfaces between the superimposed models. The VAR protocol showed no statistical differences in the root mean square between different tooth segments with an overall tendency for inadequate attachment removal. No difference between the groups were found when regarding the number of attachments was used as a main factor. The VAR technique is precise enough for the fabrication of retainers from printed dental models in a clinical setting.