In order to reduce the effects of the low strength and high compressibility of soft soil, geosynthetic-reinforced pile foundations (GRPF) are widely applied for the construction of high-speed railways. Though its reinforcement effect is proved acceptable in practices so far, it is unclear whether it will keep this performance as the train speed continues increasing. Since it is impossible to study the problem in field tests, only mathematical and physical models can be used. However, the nonlinear behaviour of the soft soil complicates the use of analytical models. Therefore, this paper presents a small-scale model test to study the possible changes in stress distribution and deformation in the GRPF under increasing dynamic loads. One test with a natural foundation, without piles or geosynthetic, shows the difference with a similar construction with GRPF foundation. Furthermore, three GRPF tests show the influence of the embankment thickness. The results show the long-term dynamic loading significantly affects the dynamic stress and displacements of the subsoil between the piles of the GRPF. This effect can be divided into three stages with an increasing level of load amplitude: no impact, advantageous impact, and adverse impact. When the dynamic load reaches the adverse impact stage, the long-term dynamic loads reduce the dynamic pile–soil stress ratio, which means that more soil settlement will develop, because more dynamic stress is applied to the soft soil. The test results show that the reduction in dynamic stress on the subsoil in the GRPF construction is clearly lower than the dynamic stress on the natural foundation, due to the existence of rigid piles. Moreover, a thicker embankment gives significantly lower dynamic stresses on the subsoil between the piles. For the thickest embankment tested, the adverse impact stage was not found at all: the arching kept enhancing under long-term dynamic loading with high load amplitudes.