The increasing accuracy of 3D velocity models developed recently for Mt. Etna has enabled their use today in routine earthquake locations. In this work, we tested the potential and performance of a global‐search probabilistic earthquake location method (NonLinLoc) in a 3D velocity model, to improve earthquake locations for seismic surveillance. In addition, NonLinLoc hypocenter locations and those obtained by standard iterative‐linear 3D locations, SimulPS‐14, have also been compared. To this end, a dataset of 328 selected earthquakes, occurring during the 2002–2003 Etna flank eruption, and the recent highest resolution 3D velocity model, have been used. The results revealed that the differences in hypocentral coordinates between the two methods are typically of the same order or smaller than the spatial location uncertainty. To evaluate the consistency of results between the two 3D location algorithms, synthetic datasets with real source–receiver configuration are also considered. Furthermore, by using NonLinLoc we estimated the influence of the source–receiver geometry on the quality of hypocenter locations. If we vary the network geometry in a dense and well‐distributed network like at Etna, reducing the number of stations (by 20% and 50%), it is significant that no large systematic hypocentral shifts of the relocated earthquakes are observed if they occur within the network. NonLinLoc is a fast and promising approach for automatic earthquake locations and surveillance purposes at Mt. Etna, because (1) it works well with a reduced number of seismic pickings, which are usually available in the automatic locations; (2) it is not particularly sensitive to tolerable levels of random noise in arrival times; and (3) it produces full location uncertainty and resolution information with respect to standard iterative‐linear 3D locations.