The slip between steel and concrete components and the bond strength between the two materials are important factors in load transfer under fire and cooling conditions. This study investigates the effect of the post-heating interaction of concrete and steel on the performance of composite members considering the bond between the two materials. Unidirectional push-out tests were conducted on circular concrete-filled stiffened steel tube (CFSST) columns after experiencing heating and subsequent cooling. The parameters under study were the diameter-to-thickness ratio of the tube (63.5, 31.75, and 21.17), the number of shear connectors (0, 4, and 6), and the target temperature (25, 250, 500, and 750 °C). The results showed that shear connectors effectively prevent concrete from sliding against the steel surface, facilitating load transfer from concrete to the steel tube. This increased interlocking between steel and concrete surfaces, leading to a substantial improvement in bond stiffness by 1–284% and bond strength by 2–590% for specimens with 4 and 6 shear studs. Moreover, the reduction in the diameter-to-thickness ratio enhanced interlocking between the steel tube and concrete core, leading to an increase in bond stiffness by 4–201% and bond strength by 10–216%. The application of thermal loading and the subsequent cooling induced macro-changes in the geometry of both the concrete core and steel tube, resulting in significant improvements in the bond properties of CFSST columns. In the specimens without shear connectors, slip between the concrete and steel occurred without any obstacle, while in the specimens with shear connectors, failure occurred in the form of the local crippling of the tube and the crushing of concrete around the shear stud together with the outward local buckling of the steel tube around the shear stud. Ultimately, a model to capture the bond stress-slip relationship of CFSST columns after thermal loading was proposed and then evaluated against the experimental data of the present study, which showed a good correlation between the experiment and prediction results.
