Drivers of willingness to communicate with generative AI: the roles of self-efficacy, grit, speaking enjoyment, and anxiety from a self-determination theory perspective.

Introduction: English-speaking proficiency is essential for the personal and academic development of English as a Foreign Language (EFL) learners; however, many students demonstrate limited willingness to communicate (WTC) in classroom settings. Although generative artificial intelligence (GenAI) offers considerable potential for enhancing communicative engagement, existing research has predominantly examined WTC in human-to-human interactions, leaving the applicability of established models in AI-mediated environments uncertain. Method: To address this gap, the present study, guided by self-determination theory, explores the structural relationships among learning grit, speaking self-efficacy, speaking enjoyment, speaking anxiety, and WTC with GenAI. Questionnaire data were collected from 350 Chinese undergraduate EFL learners who practiced oral English using GenAI, and were analyzed using confirmatory factor analysis (CFA) and structural equation modeling (SEM). Results and discussion: Findings indicate a robust nomological network, with speaking self-efficacy identified as the most influential direct and mediating predictor of WTC with GenAI. Learning grit exerts both direct and indirect effects, while speaking enjoyment and anxiety have comparatively modest impacts. These results suggest that self-efficacy and grit are fundamental psychological drivers of communicative action in GenAI contexts, whereas affective states play a supplementary role. The study extends current WTC frameworks to technology-mediated settings and highlights pedagogical implications for fostering self-belief and perseverance in digital language education.