Transcranial Magnetic Stimulation (TMS) established itself as a powerful technique for probing and treating the human brain. Major technological evolutions, such as neuronavigation and robotized systems, have continuously increased the spatial reliability and reproducibility of TMS, by minimizing the influence of human and experimental factors. However, there is still a lack of efficient setup procedure, which prevents the automation of TMS protocols. For example, the setup procedure for defining the stimulation intensity specific to each subject is classically done manually by experienced practitioners, by assessing the motor cortical excitability level over the motor hotspot (HS) of a targeted muscle. This is time-consuming and introduces experimental variability. Therefore, we developed a probabilistic Bayesian model (AutoHS) that automatically identifies the HS position. Using virtual and real experiments, we compared the efficacy of the manual and automated procedures. AutoHS appeared to be more reproducible, faster, and at least as reliable as classical manual procedures. By combining AutoHS with robotized TMS and automated motor threshold estimation methods, our approach constitutes the first fully automated setup procedure for TMS protocols. The use of this procedure decreases inter-experimenter variability while facilitating the handling of TMS protocols used for research and clinical routine. Highlights • Automatized setup procedures would facilitate TMS experiments and increase reproducibility • We developed a Bayesian model aiming at automatically finding the motor hotspot • Implementation of this model in a robotized TMS system allows the automatic search for the motor hotspot