With the development of bridge crossings over rivers, the accident of the vessel–bridge collision is increasing as well. It is important to assess probability of bridges colliding with passing ships. Firstly, the AIS (Automatic identify system) data was collected and decoded to obtain the dynamic information of the ships passing the bridge including the distributions of ships position, speed, and yaw angle, which are then compared with the value recommended by the AASHTO (American Association of State Highway and Transportation Officials) specification. The mainly influential parameters of ship–bridge collision obtained from AIS data are used to correct the variables in the risk assessment of AASHTO specification, which intends to improve the assessment accuracy by considering the actual information of passing vessels. The collision probability with and without considering the actual situations of passing ships are compared. It is found that the distribution and transit path of passing ships significantly influence the collision probability. To improve the risk assessment accuracy, it is suggested to use the actual distributions of passing ships from AIS data.