In the shaking table test of a soil–pile–bridge structure system, it is difficult to keep the similarity relations of the model structure and that of the model soil consistent. Due to the difference of geometry and material similarity ratios for the model structure and model soil, the determination of the duration compression ratio of input motions is a key problem. The spectrum characteristics of input motions will be varied by the duration compression ratio so that the seismic responses of structure and soil system will be affected. There are three commonly used approaches to determine the duration compression ratio of input motions in shaking table tests: the time similarity ratio of model structure; the time similarity ratio of model soil; and uncompressed. To study the influence of the duration compression ratio on the seismic response of a soil–structure system in shaking table tests, the El Centro record and the Wolong record were chosen as the input motions, and the durations were compressed by the three commonly used approaches in this paper. The influence of the duration compression ratios of the input motions on the acceleration response of a soil–pile–bridge structure system was compared and analyzed through a series of shaking table tests. The results showed that the duration compression ratio affected the acceleration response of the model soil and the model structure, and the effect was more obvious when the peak ground acceleration (PGA) was small. If the research is focused on the seismic response of the soil, it is recommended to use the time similarity ratio of the model soil to compress the input motions. If the research is focused on the seismic response of the structure, it is recommended to use the time similarity ratio of the model structure to compress the input motions. This study could provide a reference for the design of the shaking table test of a soil–pile–bridge structure system.