Interactive reading patterns and metacognitive strategies from surface to deep levels of L2 learners in GenAI-assisted critical reading

Engaging critically with academic literature is challenging yet indispensable for research-oriented postgraduate students. Although GenAI-powered tools have been developed to summarize, evaluate, and revise academic texts, there remains a paucity of research on how these tools specifically support the reading processes of second language (L2) readers. To address this gap, this study recruited 53 postgraduate students who were tasked with drafting and revising a critical review of a published academic article using three resources: (1) critical reviews generated by a GenAI-powered tool developed on the Coze platform, (2) communications with an embedded GenAI chatbot, and (3) the original academic article. Lag Sequential Analysis (LSA) revealed three distinct patterns of learner interaction: (1) predominantly relying on GenAI-generated critical reviews with minimal rereading of the original academic article; (2) primarily engaging in careful reading of GenAI-generated critical reviews, supplemented by frequent rereading of the original academic article; and (3) extensively integrating both GenAI-generated critical reviews and direct chatbot interactions. Qualitative interviews further indicated that these interaction patterns were influenced by learners’ deployment of metacognitive strategies adapted to task complexity and cognitive demands. Specifically, Pattern 1 was characterized by surface-level strategies for less challenging tasks, whereas Patterns 2 and 3 involved deeper, more extensive strategies: exploring, evaluating, and synthesizing ideas across GenAI-generated critical reviews, chatbot interactions, and the original academic article. These findings suggest that instructors should scaffold students away from sole reliance on GenAI-generated content toward multi-source evaluation to foster deeper metacognitive strategies and enhance reading depth.