Engineering structures are often subjected to dynamic loads, which have negative influence on the structures. In order to accurately and effectively detect the load states of structures, a cuckoo search (CS) algorithm based on the Newmark-β method is developed to identify dynamic loads. Firstly, the time is discretized into several time steps, and the discretization equation of motion is obtained. The motion equation of the discrete motion system is solved and the response of the structure under dynamic load is calculated by the Newmark-β method. Secondly, the dynamic response is selected as the optimization variable. The objective function of the dynamic load identification problem is defined as the difference between the calculated response and the measured response of the dynamic load. The dynamic load is determined through minimization of the objective function with the CS algorithm. Finally, a simply supported beam bridge subjected to dynamic load is taken as an example to verify the accuracy and effectiveness of the algorithm. The effects of the nest number, the locations of measurement points, the number of measurement points and the measurement noise on the inversed results are discussed in numerical examples. The results show that the CS algorithm can be an accurate and effective inverse system method for dynamic load identification.