Concrete-filled glass fibre-reinforced polymer tubes (CFFTs) have reportedly been considered for industrial applications such as fender piling for marine structures and bridge girders. Even though previous investigations of the flexural fatigue behaviour of circular sectioned CFFTs have been reported in the literature, no such study has been carried out for square or rectangular sectioned tubes. In order to address this research gap, an experimental programme was carried out to investigate the flexural fatigue behaviour of square sectioned CFFTs, considering fatigue deformation cycles equal to 75%, 80%, 85% and 90% of the deformation corresponding to the static flexural failure of the tested CFFTs. The used glass fibre-reinforced polymer (GFRP) tubes contained reinforcements at the longitudinal direction and ±45° to the longitudinal axis of tubes. They were also specified to have much larger longitudinal tensile and compressive strength compared to typical rectangular GFRP tubes. Self-compacting concrete was used for the core-infill of the tested CFFTs. In total, eight static flexural tests and sixteen flexural fatigue tests were conducted under displacement-control fatigue loading using two loading arrangements, namely three-point and four-point bending. All CFFTs beams subjected to the fatigue loading failed due to buckling of the compression flange of CFFT. It was observed that the number of cycles to failure for the employed fatigue deformation ranges were quite low and decreased with increasing the amplitude of fatigue deformation. The bending stiffness of the tested specimens was observed to decrease as the fatigue loading progressed. It was also found that the amount of dissipated energy in a given cycle and the rate of stiffness degradation increases with an increase in the amplitude of fatigue deformation, especially for specimens tested under four-point bending. Furthermore, the bending stiffness at failure for the specimens tested under four-point bending was found to be approximately 80–90% of its initial value.