Bridges are vulnerable to the fatigue damage accumulation caused by traffic loading over the service period. A continuous growth in both the vehicle weight and the traffic volume may cause a safety hazard to existing bridges. This study presented a computational framework for probabilistic modeling of the fatigue damage accumulation of short to medium span bridges under actual traffic loading. Stochastic truck-load models were simulated based on site-specific weigh-in-motion measurements. A response surface method was utilized to substitute the time-consuming finite element model for an efficient computation. A case study of a simply supported bridge demonstrated the effectiveness of the computational framework. Numerical results show that the simulated fatigue stress spectrum captures the probability density functions of the heavy traffic loading. The equivalent fatigue stress range increases mostly linearly in the good road roughness condition with the growth of the gross vehicle weight. The vehicle type and the road roughness condition affect the stress range. The influence of the driving speed on the equivalent stress range is non-monotonic. The bridge fatigue reliability has a considerable increase even under a relatively high overload limit. It is anticipated that the proposed computational framework can be applied for more types of bridges.