Would you share secrets with a chatbot? A longitudinal study on the relationship between virtual companionship and privacy disclosure

Purpose This study aims to investigate the relationship between virtual companionship (VCF) and privacy disclosure (PD) behavior in human-chatbot interactions. It aims to understand how the frequency of chatbot use, perceived value (PV), and perceived risk (PR) influence users’ willingness to disclose personal information. The research extends the privacy calculus theory into AI-based VCF contexts, providing insights into how users balance risks and benefits when interacting with chatbots. Design/methodology/approach A 3-wave longitudinal study was conducted from May 2023 to May 2024, involving 657 university students. The study employed structural equation modeling to explore the relationships between the frequency of VCF, PV, PR and PD. The mediating effects of PV and PR were also tested to offer a comprehensive view of the privacy calculus in AI-chatbot interactions. Findings The results show a significant positive correlation between the frequency of VCF and PD. PV positively mediated this relationship, suggesting that users disclose more information when they perceive higher benefits from the interaction. Conversely, PR negatively mediated the relationship, indicating that lower perceived risks lead to greater PD. Overall, frequent chatbot interactions increase PD, with users perceiving greater value that offsets potential privacy concerns. Originality/value This study is one of the first to apply privacy calculus theory to the context of VCF with artificial intelligence (AI) chatbots. It highlights the unique dynamics of PD in AI-driven interactions, offering valuable insights for chatbot developers to enhance user trust and manage privacy risks. The findings provide a framework for understanding how users balance risks and benefits when forming emotional connections with AI, which is crucial for future AI design and privacy policy development.