Interest-Driven Search in AI-Mediated Information Environments: An Audio Diary Study

The shift from traditional search engines to LLM-based conversational AI systems has transformed how people interact with information. However, the role of AI systems in supporting interest-driven rather than goal-directed searches remains underexplored. Through a one-week naturalistic study with 19 participants and 95 audio diaries, we examine interest-driven search, in which search is initiated and sustained by user interest, and the direction and depth of how exploration shifts as that interest develops. We present the AI-Mediated Interest-Driven Search (AIM-IDS) model, which reveals how phases of interest development correspond to search behaviors in AI-mediated environments. Our findings show that while AI-mediated search lowers barriers to initiating exploration, the same features that support early-phase interest development can hinder sustained engagement in later phases. We conclude with design implications for AI systems that support interest development across different interest types.