To address the issues of external disturbances and nonlinearity in an anti-aircraft gun servo system, this paper proposes an enhanced active disturbance rejection controller based on a genetic algorithm (GA)-optimized BP neural network (GA-BPNN-ADRC). To overcome the local optima and overfitting problems encountered when tuning ADRC gain coefficients using a BP neural network alone, a genetic algorithm is introduced to optimize the initial weights of the neural network, thereby improving its global search capability. Simulation results demonstrate that under this control strategy, the servo system achieves a steady-state step response error of 0.015° and a maximum sinusoidal tracking error of only 0.064°. Compared with conventional PID and standard ADRC methods, the proposed GA-BPNN-ADRC exhibits superior dynamic response speed and tracking accuracy.