Cellular network operators have as main objectives the personalization of the offered services to their subscribers in an individual way or for groups with the same preferences. In this paper a new Personalized Recommendation System Architecture based on the geographic localization of the cellular users is described over which different services or solutions can be offered. For the mechanism of the users data collection, the Cell Broadcast technology is used, forming groups of users with similar preferences according association algorithms. This system also considers different presence levels (available, busy and absent) of the cellular users as an additional decision factor for sending or not the recommendation message and using for this purpose a application server connected to IMS (IP Multimedia Subsystem). A feedback mechanism based on the user satisfaction degree is also considered, to improve the efficiency of the system. © 2010 IEEE.