To address the error degradation of fiber-optic gyroscopes (FOGs) under high-dynamic conditions, which adversely impacts their performance, this study proposes an adaptive phase-modulation based dynamic error suppression method. Firstly, the principal analysis and system modeling of the FOG closed-loop operating system are conducted to derive the open-loop and closed-loop transfer functions, revealing significant tracking errors for angular acceleration inputs. Secondly, the relationships between the output signal noise, bandwidth of the FOG, and phase modulation depth are analyzed using the random walk coefficient. Subsequently, an adaptive relationship model between phase modulation depth and angular acceleration is established, and a simulation model of the FOG closed-loop operating system is constructed based on this model to validate and analyze the adaptive phase modulation method through simulation. Finally, the adaptive phase modulation method is implemented on a FOG prototype according to the established model and compared with the traditional fixed phase modulation method. Static and dynamic performance tests and analyses are conducted on the FOG prototype under these two different phase modulation methods. The experimental results demonstrate that both the adaptive and traditional phase modulation methods yield a random walk coefficient of 0.006°/h, while the adaptive method reduces the tracking error for angular acceleration inputs by approximately 66.73%. This approach not only effectively enhances the dynamic performance of the FOG while maintaining its static performance but also meets the bandwidth requirements under various dynamic conditions, offering significant theoretical guidance and engineering application value for improving the adaptability of FOG in complex environments.