It's her! Investigating relationship development with social AI chatbots

As AI chatbots become integrated into various aspects of everyday life, users increasinglyform parasocial relationships with their digital companions. To understand how such relationshipsemerge, we examined how perceived human-likeness fosters parasocial attachment to AIchatbots. We test a relationship process model in which human-likeness—conceptualized as ameta-factor comprising anthropomorphism, authenticity, and social presence—promotesparasocial attachment via user gratifications and interaction behavior. In a preregistered survey of598 chatbot users from 64 countries, structural equation modeling showed that perceived humanlikenesswas a strong predictor of parasocial attachment. As predicted, this effect was partiallymediated by increased gratifications and interaction. Additional qualitative responses illustratedhow chatbots provide emotional support, role-play, and productivity benefits, alongside concernsabout dependence, social withdrawal, inappropriate behavior, and privacy. Together, the findingsidentify human-likeness, gratifications, and interaction as central drivers of parasocial attachmentto AI chatbots.