Toward Integrating AI Chat and Search: A User-Centered Perspective across Age Groups
Chen He, Michiel Spape, Khadijatul Kobra, Robin Welsch, Giulio Jacucci · 2026 · ACM Transactions on Interactive Intelligent Systems · doi:10.1145/3777486
Summary
This paper investigates how younger adults (25 participants, mean age 27) and older adults (22 participants, mean age 56) use AI chat (ChatGPT) and traditional search (Google Search) together to retrieve information. The research is framed around Visual Data Exploration, specifically exploring CO2 emission data and linking findings to real-world events and policies. Participants used a custom prototype called CO2 Explorer presenting a choropleth map and line chart alongside both ChatGPT and Google Search, activated by a single unified query input. Two information retrieval task types were compared: a quantity-oriented task (find as many evidence-linked data discoveries as possible) and a quality-oriented task (generate a hypothesis with rationale and external evidence). The study was conducted remotely via Zoom, with screen recordings, interaction logs, query timestamps, and questionnaire data collected and analyzed. Note quality was assessed by two independent raters using a five-point rubric, achieving strong inter-rater reliability. The paper addresses a genuine gap in the literature: while AI chat tools have been widely adopted, relatively little user research has examined how to effectively integrate them with traditional search, especially across diverse age groups. Prior studies comparing ChatGPT and Google separately have produced mixed results depending on domain and task type. This work uniquely examines both platforms simultaneously, side by side, and derives five practical design requirements for combining them in aging-friendly, trustworthy information retrieval systems.
Key findings
Younger adults spent more time on tasks, issued more queries, explored more external web pages, and performed more visual data exploration actions than older adults across both task types. Despite lower self-reported familiarity with AI chat, older adults relied more heavily on ChatGPT, performing fewer search actions and clicking fewer links. Older adults submitted longer queries during the quality task and reached their first query faster in the quantity task, consistent with a top-down information retrieval strategy leveraging crystallized knowledge. Critically, task outcomes in terms of note quality and quantity were statistically similar across age groups, suggesting AI chat effectively compensated for older adults reduced engagement with traditional search. Both age groups rated ChatGPT as more helpful than Google Search (median: 5 vs. 4), though participants valued the source authenticity that search provided. Combining chat and search appeared to improve note grades for the quantity task compared to using either tool alone. The tool received a median System Usability Scale score of 77.5. Based on the empirical findings, the authors proposed five design requirements: (1) automatic or manual routing of queries to chat or search based on user intent; (2) proactive keyword suggestions extracted from chat answers to seed searches; (3) visual cues such as coordinated highlighting to link chat responses with search results; (4) AI-powered summaries of web page content with follow-up question support; (5) logging IR provenance and enabling user control over which contexts inform AI responses.
Relevance
This paper is directly relevant to accessibility practice because it demonstrates that conversational AI interfaces can lower barriers to information retrieval for older adults, a population that often struggles with traditional keyword-based search. The finding that older adults effectively leveraged AI chat despite lower prior familiarity supports the case for conversational interfaces as an accessibility tool, not just a convenience feature. The five design requirements have practical implications for organizations building inclusive search and knowledge tools: they highlight the need to reduce cognitive load in query formulation, validate AI-generated content through search sources, and give users control over AI context. The study also surfaces important risks: over-reliance on AI chat can reduce information diversity and expose users to hallucinations, raising concerns about AI literacy and digital trust that accessibility practitioners must address. Designers should treat chatbot-integrated tools not as replacements for search but as complementary scaffolding, particularly for users with age-related cognitive differences.
Tags: information retrieval · aging · older adults · AI chat · large language models · search engines · user study · visual data exploration · cognitive accessibility