Beyond AI Usefulness: Motivation, Engagement, and Vietnamese Students’ Self-Perceived English-Speaking Competence

AI-supported English speaking offers opportunities for rehearsal and feedback, yet the relationship between perceived AI usefulness and learners’ self-perceived communicative competence remains insufficiently understood. This cross-sectional study examined associations among perceived usefulness, intrinsic motivation, higher-order engagement, deep and surface learning, and self-perceived communicative competence. Survey data from 420 Vietnamese undergraduate and postgraduate students with experience in AI-supported English speaking were analysed using partial least squares structural equation modelling. Engagement was conceptualised as a reflective higher-order construct comprising behavioural, cognitive, and agentic dimensions. The findings revealed positive associations between perceived usefulness and intrinsic motivation, as well as between intrinsic motivation and engagement. Engagement was positively associated with deep learning and self-perceived communicative competence (β = .460), but negatively associated with surface learning. Deep and surface learning also showed significant associations with communicative self-perceptions in opposite directions, although their direct associations were weaker than that of engagement. The model explained 35.0% of the variance in self-perceived communicative competence, rather than objectively assessed speaking proficiency. These findings highlight the importance of purposeful interaction, critical evaluation of AI feedback, and learner-initiated clarification in fostering meaningful English-speaking practice beyond AI use alone. Longitudinal studies incorporating direct speaking assessments are needed to determine whether these associations extend to actual communicative performance.