The double-edged sword effect of generative AI anthropomorphism on users’ emotional attachment: the moderating role of task types

Purpose With the rapid advancement of generative artificial intelligence (AI) technologies, anthropomorphic design has garnered increasing attention for its potential to enhance user experience and facilitate more natural human–AI interaction. However, the underlying mechanisms through which anthropomorphism in generative AI influences users’ emotional attachment remain insufficiently understood. Design/methodology/approach Drawing on the stimulus-organism-response (S-O-R) framework, this study designs two experiments to test eight proposed hypotheses. We first explore the dual pathways-enabling and inhibiting through which generative AI anthropomorphism affects users’ emotional attachment and subsequently analyze the moderating role of task types in the effect of generative AI anthropomorphism on users’ emotional attachment. Findings The results show that social presence and identity threat play dual mediating roles in generative AI anthropomorphism, influencing users’ emotional attachment. Specifically, generative AI anthropomorphism enhances users’ emotional attachment by increasing social presence, while it weakens emotional attachment by triggering identity threats. In addition, task types moderate the relationship between generative AI anthropomorphism and social presence and identity threat. Originality/value This study elucidates the complex mechanisms linking generative AI anthropomorphism and user emotional attachment within varying task types. The findings offer practical insights for designers and developers aiming to calibrate the degree of anthropomorphism in generative AI to optimize user experience and engagement.