In cognitive neuroscience, optically tracked frameless stereotaxic navigation has been successfully used to precisely guide transcranial magnetic stimulation (TMS) to desired cortical areas for brain-mapping purposes. Thereby, potential sources of imprecision are the fixation of a reference frame to the head of the subject and the referencing procedure according to certain landmarks (LM). The aim of our study was to evaluate the accuracy of frameless stereotaxic coil positioning in a standard experimental setting. A parameter for accuracy is the reproducibility of LM coordinates. In order to test the stability of the referencing for stereotaxic positioning within a single TMS session (within-session stability), the coordinates of six predefined facial LM in nine subjects were recorded first after the initial registration and second after a 20 minutes TMS session. The two sets of coordinates were then compared. The reliability of the positioning coordinates between different TMS sessions (inter-session repeatability) was addressed by registering the subjects LM coordinates in two independent TMS sessions. The variance of the recorded coordinates was analyzed. Altogether, LM were registered 1728 times (192 measures per subject). Within-session stability: The mean Euclidean distance (MED) between the LM position coordinates before and after a TMS session was 1.6 mm, when pooling over all LM. Inter-session repeatability: The MED between the LM positions recorded after the reference procedures of two different sessions showed an average deviation of 2.5 mm. In conclusion, optically tracked frameless stereotaxic coil positioning is from the technical viewpoint of high stability and repeatability. It is therefore a precise method for TMS brain mapping studies or for repeated TMS treatments, with the need of topographically exact stimulation.