Transport networks rely on well-maintained physical infrastructure to allow them to operate efficiently. Bridges are particularly important, and the failure of a bridge can cause major disruption to the network. Current practices for inspection and monitoring of bridges are slow and expensive, meaning that it is not feasible to constantly monitor the vast quantities of bridges on a transport network. This paper proposes an approach which leverages data measured from in-vehicle sensors to monitor the condition of bridges, without requiring any sensors to be installed on the bridges. The proposed approach uses a machine learning algorithm to account for the influence of varying vehicle speed, and experimental tests show that changes in the structural behavior of a bridge can be detected from measurements taken on the passing vehicle. This approach represents a scalable solution for network-level bridge condition monitoring, which could be extended to account for the effects of various environmental or operational factors which may influence the measured vibrations in a full-scale scenario.