Although generative artificial intelligence (GenAI) is reshaping language education globally, its integration into English for Specific Purposes (ESP) instruction in Iran has remained limited by rigid curricula, restricted autonomy, and lack of authentic, discipline-specific resources. This study sought to address this gap by developing a contextualized co-design framework to align AI affordances with local pedagogical realities. Guided by theoretical triangulation, a hybrid methodology was employed that combined design-based research, co-design, and participatory qualitative approaches. The needs analysis phase involved twenty-six ESP stakeholders, including eight instructors and eighteen learners. In-depth interviews were then conducted with a subset of nineteen participants—eight instructors and eleven learners. Three instructors and three learners, chosen purposively among those involved in both prior phases, along with three policymakers and three AI experts, participated in the last co-design phase to ensure multiple perspectives in model development. As revealed by the findings, the learners prioritized personalization, writing support, and disciplinary adaptability, whereas the instructors emphasized controlled prompting, ethical literacy, and assessment redesign. Concerns shared between the two groups included overreliance, epistemic authority, and unequal digital access. The co-design process generated a cyclical instructional model incorporating dual human-check mechanisms, ethical self-reporting, bias-awareness checkpoints, multimodal feedback loops, and institutionalized teacher training. Offering a theoretically grounded and adaptable reference for discipline-specific AI use in ESP programs, this study’s framework can guide educators, curriculum designers, and policymakers in ESP contexts analogous to those investigated in this research.
Masoud Azadnia (Tue,) studied this question.