This study aimed to validate the Emotional Skills and Competence Questionnaire (ESCQ-42, Takšić et al., 2009) in Croatian and Portuguese samples using exploratory graph analysis (EGA). The ESCQ, rooted in the Mayer and Salovey emotional intelligence model (Mayer & Salovey, 1997), was administered to 627 Portuguese students (Mage = 15.5; SD = 0.76) and 562 Croatian students (Mage = 16.3; SD = 1.07). This questionnaire, featuring three subscales for perceiving, expressing, and managing emotions, has shown strong psychometric reliability across various studies and countries. Despite its widespread use and validation, item-level inconsistencies in cross-cultural research motivated the use of EGA. The EGA revealed that the Croatian sample exhibited a five-dimensional structure, while the Portuguese sample presented a four-dimensional structure, thus challenging the conventional three-factor model. Additional analyses of item stability, replicability, and network loadings provided deeper insights into item placements and contributions. Model fit comparisons between EGA-identified structures and the theoretical three-factor model were conducted using confirmatory factor analysis (CFA). Furthermore, EGA confirmed the three-factor structure on an abbreviated scale, achieving both configural and partial metric invariance across cultures. These findings underscore the necessity for ongoing refinement of the ESCQ-42 to enhance its cross-cultural applicability and reliability.