["European Journal of Education, Volume 61, Issue 2, June 2026. ", "\nABSTRACT\nThis study examined how Chinese EFL teachers' teaching enjoyment, teaching efficacy, and AI integration (particularly ChatGPT) relate to task‐based language teaching (TBLT) implementation. Survey data were collected from 360 teachers reporting the use of the TBLT approach. Ordinary least squares regressions with HC3 robust standard errors yielded three key findings. First, teaching enjoyment and AI integration independently predicted TBLT adoption, with personal enjoyment and AI‐related enjoyment emerging as the strongest affective dimensions. Second, teaching efficacy and AI integration were positively associated with learner‐centred, TBLT‐informed practices; follow‐up analyses suggested that efficacy, especially engagement‐related efficacy, was more strongly linked to practice at moderate and higher levels of AI integration. Third, AI integration was positively associated with teachers' confidence in implementing TBLT, though modestly, and this relationship was moderated by training: AI use predicted higher confidence among teachers with formal TBLT training but lower confidence among those uncertain about their training. Overall, the findings indicate that AI integration supports TBLT implementation in ways that depend on teachers' pedagogical preparation, underscoring the need for sustained professional development to promote principled AI use in language classrooms.\n"]