Modal coupled effects concealed in the semblance of flutter dominated by torsional modes have an uncertain impact on the flutter characteristics of long-span bridges, making it challenging to fully comprehend the underlying physics of three-dimensional flutter. Consequently, this research has investigated the composition of flutter modality and its related modal effect mechanism through experimental and theoretical analyses. Initially, the actual flutter modality and its compositions were ascertained through full-bridge aeroelastic tests on four suspension bridges with main spans exceeding 1750 m. Subsequently, the practical modality-driven flutter analysis was developed to measure the impact of each natural mode, particularly the higher-order mode, on the flutter performance through modal-generated aerodynamic damping. Next, the genetic algorithm was introduced in this method to enhance the convergence robustness. Finally, a systematic investigation was conducted to examine the evolution of modal effects with changes in the main spans. The results of full-bridge aeroelastic tests indicate that the flutter pattern is predominantly formed by the first-order symmetric torsional mode, the first-order symmetric heaving mode, and the second-order symmetric heaving mode. In terms of modal participation and modal aerodynamic damping contribution, the second-order mode governs the flutter heaving modality but has a smaller impact on the flutter performance compared to the first-order mode. The first-order torsional mode and the first-order heaving mode provide continuously increasing positive damping and continuously decreasing negative damping respectively, while the aerodynamic damping provided by the second-order heaving mode exhibits a turning phenomenon as wind velocities increase due to the system frequency approaching the second-order natural frequency. The genetic algorithm can effectively address potential occurrences of non-convergence or switching phenomena in numerical calculations.