As a critical characteristic curve of bridge structures, the influence line (IL) is effective in evaluating the bridge damage and bearing capacity. Contrary to conventional methods using static loading to obtain IL, this study proposes a methodology to determine the rotation influence line (RIL) based on the vehicle–bridge interaction (VBI) under random traffic loads. A key feature is the employment of the machine learning algorithm to determine the RIL from the upper and lower bridge response data induced by traffic flow, which is simulated using the cellular automata (CA) approach. Subsequently, by establishing a relationship between the RIL difference and the structural damage coefficient, this study identifies the bridge damage location and degree effectively. The numerical example results validate the accuracy in identifying both single- and multi-damage cases. The parameter analysis indicates that larger vehicle speed induces higher errors in identifying the RIL and the bridge damage.