Beyond Access: How GenAI Literacy Relates to Task Performance, Motivation, Cognitive Load and Interaction Patterns in AI ‐Assisted Learning Environment

ABSTRACT Background Generative AI is rapidly being integrated into educational settings, yet equitable benefits depend heavily on students' GenAI literacy. While access to these tools is increasing, empirical evidence remains limited on how specific dimensions of assessed GenAI literacy relate to students' task performance, cognitive experiences, and actual interactional behaviours during learning. Objectives This study examined the associations of overall GenAI literacy and its four dimensions (Know and Understand, Use and Apply, Evaluate and Create, Ethics) with students' abstract‐writing performance, self‐efficacy, motivation, and cognitive load. It also explored how assessed literacy is reflected in the configurations of students' interactions with GenAI. Methods A total of 162 university students completed an abstract‐writing task using a custom GenAI assistant. Ordinary least‐squares regression models evaluated the continuous GenAI literacy scores against writing performance, motivation, and cognitive load, adjusting for prior academic writing knowledge and English proficiency. Additionally, Epistemic Network Analysis (ENA) was applied to conversational trace data to compare the configurations of questioning and commanding behaviours between higher‐ and lower‐literacy students. Results and Conclusions Higher overall GenAI literacy was positively associated with better writing performance and negatively associated with perceived mental load. Dimension‐level analyses revealed distinct roles: the Use and Apply dimension was independently associated with higher writing performance, while the Know and Understand dimension was uniquely associated with lower mental load. Neither the total score nor the sub‐dimensions showed reliable associations with motivation or self‐efficacy. Furthermore, ENA profiles demonstrated that higher‐literacy students engaged in a refinement‐oriented process, whereas lower‐literacy students relied predominantly on direct content generation and factual retrieval. These findings suggest that instruction should address both conceptual understanding and practical application of GenAI, helping students critically evaluate and refine AI‐generated content during writing.