<![CDATA[Why do students who understand effective presentation practices continue to rely on reading from slides? This study investigates the cognitive, affective, and strategic mechanisms underlying this discrepancy in EFL academic presentations. Using a convergent parallel mixed-methods design, data were collected from 16 Indonesian undergraduate students through Likert-scale questionnaires and semi-structured interviews. The findings generate three key contributions. First, text dependency is reconceptualized not as a performance deficit but as a performance-stabilization strategy, enabling learners to manage fluency and reduce anxiety under real-time communicative pressure. Second, the study identifies a persistent affective–cognitive disjunction, in which anxiety operates independently of content mastery and overrides metacognitive intention during performance. Third, the study documents the emergence of AI-mediated coping, where students use generative AI tools as real-time cognitive extensions during moments of uncertainty, revealing a previously untheorized form of distributed performance regulation. Taken together, these findings reconceptualize the knowing–doing gap as a multidimensional performance regulation system shaped by the dynamic interaction of cognition, affect, and technology. Building on this integration, the study proposes the PACE Framework (Psychological–Affective–Cognitive–Ecological AI) as a pedagogical model for supporting adaptive and authentic communication in AI-mediated learning environments. The study contributes to the extension of Communication Apprehension theory into digital contexts and advances a distributed cognition perspective on L2 oral performance. Pedagogically, it highlights the need for integrated approaches that address not only speaking skills, but also affective regulation, cognitive load management, and ethical AI use in academic communication.]]>
