Comparing structural and conversational AI summarisation for time-constrained academic reading support among graduate students

Academic reading can be challenging for students due to the complex structure and language of academic texts. Emerging AI-assisted approaches can facilitate academic reading, but how different AI-assisted approaches impact students requires further research. To investigate this, we conducted a user study (n = 29) where participants completed paper categorisation tasks under three conditions: (1) with conversational AI summarisation, (2) with structural AI summarisation, and (3) without AI support. Our findings indicate that structural AI summarisation helps students navigate text structure and locate information, while demonstrating a more positive user experience and lower perceived cognitive load compared to conversational AI summarisation. However, concerns emerged, including time-consuming verification and over-reliance. We build on our results to suggest design opportunities for future AI-assisted reading support tools: (1) combining structural overviews with conversational flexibility, (2) aligning summarisation with reading objectives through intent clarification, (3) embedding verification features that maintain reading flow, and (4) shifting from direct answer provision to guided questioning that promotes critical engagement. Our findings are bounded by 10-minute paper categorisation tasks performed by HCI graduate students across two LLM implementations. Our paper contributes empirical findings comparing AI summarisation approaches and suggests design opportunities that support interactive and critically engaged academic reading.