AI-associated academic writing anxiety in AI-assisted contexts: evidence from Chinese EFL postgraduate students.

Introduction: AI adoption in academic writing provides support but also introduces uncertainty for L2 learners. Research has focused on tool effectiveness and acceptance, with limited attention to how technology perceptions influence emotional experiences, particularly writing anxiety. Methods: This study proposes a Technology-Emotion-Performance (TEP) framework, integrating the Technology Acceptance Model, social cognitive theory, and appraisal-based emotion theory. Technology appraisals (perceived usefulness, ease of use, trust), writing self-efficacy (perceived control), and writing anxiety (appraisal-based outcome) were examined. Survey data were collected from Chinese EFL postgraduate students, and structural equation modeling (SEM) analyzed the relationships. Results: Positive technology perceptions were linked to higher writing self-efficacy, which predicted lower writing anxiety. Trust had both direct and indirect effects on anxiety, whereas integrity-related threats increased anxiety. Writing self-efficacy and anxiety were associated with self-reported writing performance. Discussion: Findings indicate that writing anxiety in AI-assisted contexts is better understood as a technology-mediated appraisal outcome rather than a direct effect of AI use. Integrating cognitive, affective, and behavioral dimensions offers a clearer account of learner experiences and informs pedagogical strategies for AI-supported academic writing.