Academic Research Reveals Conversational AI and Search Follow Different User Patterns
Key Takeaways
- ▸Conversational AI sessions are significantly more self-contained than search sessions (34.1% vs 19.5%), suggesting users view AI assistants as standalone resolution tools rather than entry points to the web
- ▸Search and conversational AI occupy different structural positions in user journeys: search anchors the beginning, while AI assistants sit deeper and often serve as end-points
- ▸Directional patterns differ significantly between the two modalities within the same user (+20.6 percentage points), showing fundamentally different interaction flows
Summary
A new arXiv research paper analyzing user behavior across search engines, browsing, and conversational AI assistants reveals a fundamental shift in how people seek information online—not through replacement, but through "recomposition" of search journeys. Researchers tracked user sessions across standalone AI assistants and traditional search engines, finding that conversational AI creates distinctly different interaction patterns than search. The study discovers that 34.1% of conversational AI sessions are entirely self-contained (showing no external web searches), compared to just 19.5% of search-centered sessions among the same users—a difference of 13 percentage points.
The research challenges the assumption that conversational AI is replacing search. Instead, the data shows these tools occupy different roles within user journeys. While search typically anchors the beginning of information-seeking activities, conversational AI assistants sit deeper within journeys, often serving as contained stopping points. Session patterns show that only 10.5% of sessions begin with AI assistants ("AI-first"), while 18.3% end with them ("AI-last"), and 37.1% interleave or bridge both. Within individual users, the directional pattern between search and assistant use differs by 20.6 percentage points, indicating these are genuinely different tools serving different functions.
- Only 10.5% of sessions are AI-first, indicating most information journeys still begin with traditional search rather than jumping directly to conversational interfaces
Editorial Opinion
This research provides empirical evidence that conversational AI and search are co-evolving as complementary tools rather than competing technologies. For companies building search and AI products, the implication is clear: users view these as serving distinct functions within their information-seeking workflows. Rather than AI assistants cannibalizing search traffic, the research suggests opportunities for more sophisticated cross-platform experiences that respect these functional differences while enabling users to move fluidly between modalities.
