Rethinking search intent: From traditional search engines to LLM-powered information retrieval

Abstract

This study examines the limitations of traditional search intent frameworks in the evolving landscape of digital information retrieval. We argue that Broder's (2002) widely used taxonomy of informational, navigational, and transactional intent fails to capture contemporary search behaviors due to two major shifts: first, traditional search engines have evolved from simple hyperlink providers into platforms offering rich snippets and direct answers; second, large language model (LLM) chatbots have emerged as alternative information retrieval tools. Through an online survey we conducted (N=82) where participants reflected on their actual search histories across both platforms (246 sessions each), we identified three key shortcomings of Broder’s (2002) taxonomy: emergent patterns outside existing categories, dissolution of boundaries between intent types, and statistically uneven distribution of categories across platforms. Based on our analysis of participants' reflections on their search histories and the identified shortcomings of traditional search intent taxonomies, we propose a novel user-centered framework. This framework shifts the focus from what users search for to why they search and how they use information. Our model has the following dimensions: immediate search goals (knowledge-oriented, solution-oriented, or resource-oriented), contextual triggers, outcome realizations, and overarching purposes connected to fundamental human values. This approach, grounded in Xie's (2002) model of interactive information retrieval and Schwartz's Theory of Basic Values (2012), provides a more comprehensive understanding of search behavior by considering the entire search journey rather than isolated queries, offering valuable insights for designing more responsive information retrieval systems.