We present a geolocation method for GSM/CDMA/UMTS networks. A cell area is partitioned into a grid, and a sequential Bayesian updating scheme is proposed to identify the grid point within the circular belt, defined by the one-way delay between the mobile and the base station, that is the most likely position of the user. Using GSM as case study, we employ an RF simulation model to study the accuracy of the algorithm compared to existing methods. We show that the accuracy of the Bayesian update method is relatively insensitive to cell size and robust to parameter settings.