Bonding with AI chatbots: anthropomorphism and changes in social support networks using PLS-SEM and fsQCA

Purpose Artificial intelligence (AI) chatbots have become increasingly embedded in everyday communication, serving not only as utilitarian tools but also as potential sources of social and emotional support. However, research has largely focused on adoption patterns and interaction outcomes, with limited attention paid to how AI interactions reshape users’ social support networks and the mechanisms underlying these changes. Design/methodology/approach This study develops a process-oriented model integrating individual characteristics (social interaction avoidance, loneliness, emotional dependency and self-disclosure) and AI chatbot characteristics (perceived AI response quality and perceived AI credibility). Data from 688 workplace users are analyzed using a mixed-methods approach combining partial least squares structural equation modeling and fuzzy-set qualitative comparative analysis. Findings Although perceived AI response quality does not exert significant direct or indirect effects, perceived AI anthropomorphism is primarily driven by individual characteristics and credibility-related cues, suggesting that functional performance may be insufficient to elicit higher levels of social–relational perceptions. Configurational analysis reveals that changes in social support networks emerge through multiple equifinal pathways shaped by different combinations of individual and AI-related conditions. Originality/value This study extends research on AI anthropomorphism by shifting the focus from interaction-level outcomes to broader social–relational consequences and by demonstrating the value of integrating symmetric and configurational approaches to capture the underlying mechanisms. It also provides practical insights for designing AI systems that balance emotional support with healthy offline social engagement. Peer review The peer review history for this article is available at: Link to the website