Vehicles generate moving dynamic loads on bridges. In most studies and current practice, the actions of vehicles are modelled as moving static loads with a dynamic increase factor, which are obtained mainly from field measurements. In this study, an evolutionary spectral method is presented to evaluate the dynamic vehicle loads on bridges due to the passage of a vehicle along a rough bridge surface at a constant speed. The vehicle–bridge interaction problem is modelled in two parts: the deterministic moving dynamic force induced by the vehicle weight, and the random interaction force induced by the road pavement roughness. Each part is calculated separately using the Runge–Kutta method and the total moving dynamic load is obtained by adding the forces from these two parts. Two different types of vehicle models are used in the numerical analysis. The effects of the road surface roughness, bridge length, vehicle speed and axle space on the dynamic vehicle loads on bridges are studied. The results show that the road surface roughness has a significant influence on the dynamic vehicle–bridge interaction. The dynamic amplification factor (DAF) and dynamic load coefficient (DLC) depend on the road surface roughness condition.