Orchestrating ChatGPT in English language teaching: a complex dynamic systems study of teacher agency

Purpose ChatGPT is used extensively in English Language Teaching (ELT), yet its effects on teachers’ agency and professional autonomy remain underexplored. Prior work focuses on student use, leaving limited evidence on how teachers navigate, adapt to, or resist AI in pedagogical work.Design/methodology/approach Using Complex Dynamic Systems Theory (CDST), we analysed teacher – AI interaction in Australian tertiary ELT. A qualitative-dominant mixed-methods design combined a survey of 30 teachers with in-depth interviews.Findings Teachers used ChatGPT routinely for time-intensive, language-explicit tasks (planning, grammar checking, discourse modelling, material design). Confidence and perceived affordances varied, and use involved iterative prompt–output feedback loops. Participants demonstrated a ‘teacher-as-editor’ role, maintaining pedagogical control and professional judgement. Prompt literacy emerged as a core competence supporting integration while sustaining autonomy. Institutional policy exerted little influence; teachers rather self-regulated based on judgement and peer exchange.Originality/value The study advances theory on digital teacher agency by showing that agency in AI-mediated ELT is not simply a matter of adoption or resistance, but a task-sensitive process of orchestration shaped by prompt-output-evaluation-revision cycles. Empirically, it identifies the teacher-as-editor as a distinctive professional role through which teachers preserve pedagogical control, and it shows how this role can inform innovation in pedagogy, prompt literacy, and institutional guidance.