AI anthropomorphism, emotional attachment and verification behaviour in GenAI use

Purpose With the rapid advancement of generative artificial intelligence (GenAI), AI is increasingly perceived not merely as a tool but as a conversational partner. This study examines how AI anthropomorphism is associated with users’ verification of AI-generated information through emotional attachment. Design/methodology/approach A survey was conducted with 430 adult GenAI users in South Korea. The proposed mediation model was analysed using structural equation modelling. Findings Within the proposed structural model, AI anthropomorphism was positively associated with emotional attachment to AI, which was in turn positively associated with AI-generated information verification behaviour. AI anthropomorphism was also directly and positively associated with verification behaviour. The significant positive indirect association along the anthropomorphism–attachment–verification pathway suggests that the findings are more consistent with a cognitive vigilance pathway than with a cognitive illusion pathway. Rather than indicating that anthropomorphism and attachment reduce scrutiny, the results suggest that they are associated with greater verification effort. Originality/value This study narrows its empirical contribution to a mediation model linking AI anthropomorphism, emotional attachment and verification behaviour. It shows that emotional attachment to AI is not necessarily associated with reduced scrutiny, but may instead be linked to greater verification effort. Building on this finding, AI verification fatigue is repositioned as a promising theoretical extension for future research on the possible longer-term cognitive burden of repeated AI-generated information verification, rather than as an empirically demonstrated outcome of the present study.