Significant increasing prevalences have been observed in gambling disorder (GD) in the last decades. This study analyzed the underlying mechanisms of the gambling severity with path analysis (implemented through Structural Equation Modeling, SEM), and assessed the potential moderator effect of the patients' sex. A sample of n = 512 treatment-seeking patients was assessed for sociodemographics and clinical state previously to the treatment. Results obtained in two separate SEM (for men and women) revealed differences in the direct effects and the mediational links. Among the male subsample, higher GD severity was directly related to the higher cognitive bias and the younger age of onset of the problematic gambling, while impulsivity levels and age of onset achieved an indirect effect on the disordered gambling mediated by the cognitive bias. Among females, GD severity was directly increased by younger age of onset, higher cognitive bias and lower self-directedness, while lower socioeconomic positions, and higher levels in harm avoidance achieved an indirect effect on the gambling severity mediated also by the distortions related to the gambling activity. These results provide new empirical evidence for a better understanding of the GD etiology, suggesting that the underlying complex links mediating the GD severity are strongly related to the patients' sex. The results can also contribute to design more effectiveness and precise therapy programs of patient-centered care.