<p>The tilt of sector antennas is a key strategic parameter for optimizing coverage and Quality of Service (QoS) in cellular networks, particularly in dense urban environments. Traditional manual adjustment methods are often costly, time-consuming, and ineffective when faced with the complexity of real physical environments. This paper introduces a hybrid approach, referred to as Hybrid AI–EM–GIS, which combines artificial intelligence techniques with an electromagnetic wave propagation simulator (EM), interfaced with a Geographic Information System (GIS) incorporating a detailed topographic model. This integration enables realistic simulation of radio performance and effectively guides the optimization process toward antenna configurations adapted to terrain constraints. Four algorithms were evaluated: Particle Swarm Optimization (PSO), Artificial Immune System (AIS), Differential Evolution (DE), and PSO-GA, a hybrid version of PSO that integrates the genetic algorithm’s mutation principle with an adaptive mutation rate. Two objective functions were considered: the optimization of the Reference Signal Received Power (RSRP), and the joint optimization of RSRP and the Signal-to-Interference-plus-Noise Ratio (SINR). The study, conducted on a representative scenario of the LTE-A cellular network operated by ATM MOBILIS and deployed northeast of the Port of Oran, shows that the PSO-GA algorithm offers the best trade-off between coverage and link quality, achieving an RSRP coverage rate of 86.3% and an average SINR of 30.0&#xa0;dB. Compared to the current operator configuration, the PSO-GA method provides a significant improvement of 62.4%. These results confirm the relevance of the Hybrid AI–EM–GIS approach for automatic antenna tilt optimization in complex urban environments, while reducing reliance on drive-test campaigns thanks to its direct compatibility with Self-Organizing Network (SON) mechanisms.</p>