Objective: This study aimed to evaluate the bond strength (BS) of glass fiber posts (GFP) at different root levels when luted with conventional or self-adhesive cements in crown-restored human premolars subjected, or not, to cyclic mechanical loading.Materials and Methods: Sixty lower premolar roots were endodontically treated and prepared for a GFP system. Half of the roots (n=30) had their posts cemented with a self-adhesive resin cement, while the remaining roots followed a three-step conditioning method: acid etch, bonding agent, and a conventional resin cement. Metal crowns were luted onto the post-core preparations and the specimens were embedded to simulate the periodontium. Half of the specimens from each group (n=15) were submitted to cyclic loading simulations (130N; 2.0Hz) and then sections were obtained from each root for the pushout BS test.Results: Independently of the cyclic loading and the root level tested, the conventional resin cement provided significantly higher values of BS (p=0.002). For either cement or either root level, cyclic loading caused a significant decrease in BS values (p=0.023). The Tukey test indicated that, regardless of the resin cement used or the cyclic loading, BS was highest at the middle and cervical thirds of the root (p=0.026), and their values did not differ between themselves.Conclusions: When used for luting GFP, self-adhesive resin cement resulted in lower pushout BS than the conventional counterpart, with cyclic loading causing a decrease in BS of the GFP to dentin for both resin cements.