Abstract: Fiber-reinforced polymer (FRP) composites, owing to their excellent specific strength and stiffness, have been widely employed in high-performance engineering fields such as aerospace, transportation, and marine industries. Under cyclic loading, FRP materials undergo fatigue damage processes characterized by multi-scale and multi-mechanism coupling, including matrix crack propagation, fiber-matrix interfacial debonding, interlaminar delamination, and fiber breakage. This complexity poses significant challenges to the accurate assessment of their fatigue performance. This review summarizes recent advances in the characterization of FRP fatigue behavior. First, key material parameters-such as fiber type, volume fraction, fiber structure (including unidirectional, two-dimensional, three-dimensional, and spread-tow braids), matrix properties, and external environmental factors such as temperature, humidity, and chemical attack-are analyzed to reveal how internal and external factors influence the damage evolution patterns and fatigue life of FRP. Second, the role of uniaxial and multiaxial fatigue testing, as well as advanced in-situ and nondestructive testing methods, in revealing the damage evolution mechanisms of FRP are summarized. Third, theoretical approaches are reviewed, including fatigue life prediction models, stiffness and strength degradation models, and progressive damage models. Finally, representative applications and emerging trends of machine learning in FRP fatigue performance prediction are discussed. By consolidating existing research, this review provides a reference framework for understanding the mechanisms of fatigue damage evolution and fatigue life prediction of FRPs, offering valuable insights into the reliability assessment of these composites under complex service conditions.