Consumers or Producers? The Feedback–Autonomy Paradox of AI Tools in Blended ESP Learning

Artificial intelligence (AI) tools now reach English for Specific Purposes (ESP) students through the smartphones they already carry, often before institutions have decided how these tools should be used. This qualitative case study examines how ESP students and teachers at a university in the Sultanate of Oman experienced AI-supported feedback in blended learning, and how the appeal of immediate feedback interacted with the need for learner autonomy. The data come from a wider doctoral study of mobile technology in ESP classes and comprise semi-structured interviews with ESP teachers, five focus groups with 30 second-year students from maritime, engineering and medical programmes, 81 open-ended questionnaire responses and three classroom observations. AI use was not a preset topic; it emerged from participants’ own accounts of the applications they used. Students described ChatGPT and AI speaking applications as an always-available, non-judgmental tutor that gave immediate feedback and helped them rehearse speaking, writing and professional communication with more confidence. Teachers welcomed this but worried that students might become consumers rather than producers of language. We interpret these accounts as a feedback–autonomy paradox: the support that builds confidence can also crowd out independent effort, while autonomy without guidance leaves learners unsure of their progress. Teachers managed the paradox through dialogical use of AI, recorded practice followed by peer or teacher feedback, and professionally authentic tasks, while cost, limited training and the absence of shared guidance constrained equitable access. We argue that the feedback–autonomy balance should be treated as a design decision that changes over the course of a programme, and we outline implications for ESP teachers and institutions.