Meaning over Quantity: An Exploratory Study on How Relationship Intimacy Shapes Giver Query Formulation in LLM-based Gift Recommendation

The rise of agentic commerce is shifting recommendation design from algorithmic precision toward contextual understanding. Despite gift-giving being a complex social ritual, how relationship intimacy shapes giver–AI interaction remains underexplored. This study exploratorily examines how intimacy influences query formation in LLM-based gift recommendation systems specifically, how givers construct prompts and convey context depending on relational closeness. Drawing on Construal Level Theory (CLT), which predicts that psychological proximity increases representational concreteness, we conducted a between-subjects experiment. Result, givers in close relationships produced contextually rich prompts elaborating recipient identity and emotional intent, while givers in distant relationships constructed sparse, norm-centered prompts minimizing personal disclosure. This study contributes to HCI by (1) characterizing intimacy-driven narrative style differences in giver queries, (2) bridging social psychology and prompt engineering, and (3) proposing a design paradigm for adjusting agent persona to relational context (Figure 1).