To foresee the life of fretting fatigue with variable amplitude loading, a hybrid model based on non-local multiaxial fatigue parameters and artificial neural networks (ANN) has been proposed. Thus, in order to compute the impact of the variable amplitude loading, the model has been developed based on equivalent stress parameters related to crack nucleation and Miner’s cumulative damage rule. To validate the models, fretting fatigue tests of several different aeronautical alloys were gathered from the literature and used in the ANN training. Furthermore, the analyzed data encompass a wide range of loading conditions and different contact geometries, which contributes to the formulation of a generalized and robust approach for the assessment of fretting fatigue problems. The proposed methodology aims to effectively address the complexities inherent to fretting fatigue under variable amplitude stress. Additionally, these types of fatigue ANN models have been demonstrating a strong capacity for generalization, making them highly promising for future industrial applications.