When AI becomes a language partner: a moderated serial mediation model of AIGC-supported communicative task engagement and learning enjoyment.

Background: Language learning is a deeply social and affective process, and much of it now unfolds beyond the classroom, through AI generated content (AIGC) tools that learners take up for communicative practice in everyday digital life. This rapid, largely unguided shift raises a timely question for the social and affective domain of language learning: whether practicing with an AI provides relational and emotional experiences that sustain language use, such as a sense of a responsive partner and the enjoyment of meaningful exchange, and how this process varies with learners' tolerance of ambiguity. Methods: A correlational survey of 780 s or foreign language learners who use AIGC tools for communicative practice tested a moderated moderated serial mediation model linking communicative task engagement to learning enjoyment through social presence and then perceived epistemic value. Ambiguity tolerance served as a moderator, and the richness of the interaction channel (text, voice, or multimodal) as a further boundary condition. Results: Communicative task engagement was positively associated with learning enjoyment, and the results were consistent with an indirect association through social presence and then perceived epistemic value. The engagement associations were stronger for learners lower in ambiguity tolerance, a compensatory pattern that was more pronounced in the richer voice or multimodal subgroup, chiefly at lower and mean ambiguity tolerance. This channel pattern was statistically reliable for the associations with social presence and with perceived epistemic value. Because the study was cross-sectional, the findings should be interpreted with caution. Conclusion: Learning enjoyment in AIGC-supported communicative practice appears to be a social and affective achievement: it is stronger when engaged learners report a sense of a responsive partner and judge the activity worthwhile, and its association with engagement varies with their comfort with uncertainty and with the richness of the channel. By centering this experience, the study contributes to the social and affective domain of language learning beyond the classroom and locates the value of practicing with AI in the relational and emotional experience it affords. Tasks that invite responsive, conversational exchange and visible learning gains may help learners, especially those uneasy with uncertainty, sustain such practice.