Tool, Assistant, or Advisor? Consumer Experience across Modalities in AI-Powered Product Search

Abstract

This study investigates how different AI-powered search interface designs—traditional keyword search, AI-generated summaries, and conversational agents—influence users’ mental representations of AI and their subsequent evaluations and behaviors. Across an experimental design involving 345 participants, we found that interface modality significantly shaped how users conceptualized the AI system's role. Cluster analyses revealed three distinct user types with divergent expectations and evaluative priorities (e.g., as a tool, assistant, or advisor). These perceived roles, in turn, moderated how traditional usability constructs—perceived usefulness, ease of use, and control—predicted decision confidence and purchase intention. These factor are weighted differently within each cluster. Our findings show that interface design actively constructs interaction realities that guide technology acceptance and consumer behavior. We discuss implications for adaptive interface design and the ontological framing of AI in user experience.