How intelligent recommendations drive green consumption: the importance of green involvement and technology self-efficacy

Purpose Drawing upon the Stimulus-Organism-Response (SOR) model, the Technology Acceptance Model (TAM), and Construal Level Theory (CLT), this study constructs a moderated mediation model to examine how intelligent recommendation-based advertisements influence green consumption intentions through empathy, while also investigating the moderating effects of green involvement and technological self-efficacy. Design/methodology/approach This study adopted an experimental design. Study 1 and 2 incorporated moderating variables using a multi-factor between-subjects design. By manipulating green involvement and measuring technological self-efficacy, the study comprehensively tested the moderated mediation model. All data were analyzed using analysis of variance (ANOVA) and the Bootstrap method. Study 3 tested the moderating role of empathy, aiming to verify that empathy serves as a more suitable mediating variable than green involvement. Findings The results indicate that intelligent recommendation-based advertisements have no direct effect on green consumption intention; instead, empathy fully mediates this relationship. Furthermore, green involvement and technological self-efficacy moderate the relationship between intelligent recommendations and empathy, thereby enhancing green consumption intention. Originality/value First, unlike previous studies focused on traditional advertising, this paper focuses on the emerging form of intelligent recommendations, revealing their unique psychological influence mechanisms. Second, it pioneers the “algorithm-empathy” pathway linking technological systems to sustainable behavior, identifying empathy as a key mediating variable. Finally, it introduces and validates two critical moderating variables – green involvement and technological self-efficacy – deepening the theoretical exploration of the boundary conditions of intelligent recommendation-based advertising.