Beyond performance: cognitive extension, affective co-regulation, and agency in AI-mediated L2 writing

As generative artificial intelligence increasingly shapes English as a foreign language writing, research has often focused on performance outcomes rather than learners’ lived experiences. This interpretive qualitative study investigates how 25 upper-intermediate EFL undergraduates in Jordan negotiate authorship, agency, emotion, and identity during AI-mediated academic writing. Drawing on reflective journals, semi-structured interviews, and ChatGPT interaction logs collected over twelve weeks, the analysis shows that learners engaged with AI as a collaborative resource rather than a corrective authority. Authorship and agency were enacted through selective adoption and resistance to AI input, emotional engagement shifted from external reassurance to greater self-regulation, and cultural identity shaped how AI-generated language was adapted to maintain authenticity within academic discourse. Interpreted through cyborg psychology as a sensitising lens, the findings conceptualise AI-mediated writing as a hybrid learning ecology, that is, a dynamic environment of learners, AI tools, classroom practices, and institutional norms that jointly shape writing processes and experiences, and offer pedagogical insights for integrating AI into EFL writing instruction in ways that support learner agency, ethical awareness, and cultural responsiveness.