Seeking With Sentiment: Emotional Attachment and the Use of Generative AI as an Information Intermediary

Generative artificial intelligence (GenAI) is rapidly emerging as a powerful intermediary for information access, reshaping how individuals seek and evaluate knowledge. While prior research has examined how people use conversational GenAI to find specific types of information and how dialogue-based search compares with traditional, non-dialogue search engines such as Google, less is known about the psychological and perceptual antecedents of this behavior. Much of the existing literature emphasizes rational factors, such as perceived information-gathering capabilities, yet affective dimensions like emotional attachment have received limited attention. Recent studies suggest that users who form emotional attachment to ChatGPT tend to rely on it for information and emotional support. Using a national US sample ( N = 900) collected by Verasight, this study examines how emotional attachment to GenAI, trust in AI, AI literacy, and risk perception influence individuals’ GenAI usage to seek various types of information. Results show that emotional attachment to GenAI is significantly associated with higher trust in GenAI, which in turn is associated with using it as an information intermediary across various domains. Additionally, the association between emotional attachment and trust in GenAI is more pronounced among individuals with low levels of AI literacy and those who perceive high risk from using GenAI for information access. In addition, AI literacy moderated the relationship between emotional attachment and health-related and political information seeking through GenAI. Findings contribute to a deeper understanding of the emotional dynamics shaping human–AI interaction and offer practical insights for the design and governance of AI systems in public information environments.