|
Abstract
|
Generative artificial intelligence (GenAI)—particularly large language model (LLM) systems such as ChatGPT and Co-pilot—has rapidly become a common learning resource in higher education. Many students now consult GenAI for explanations, examples, feedback, and drafting support, which can increase access to individualized learning help but also introduces new challenges for teaching, assessment, and academic integrity (Lund et al., 2023; Miao & Holmes, 2023; Pitts & Motamedi, 2025). Because GenAI systems can produce fluent, persuasive output regardless of whether the content is accurate, students’ learning outcomes increasingly depend not only on whether they use AI, but how they evaluate and integrate AI-generated information into their academic work (Martín-Moncunill & Alonso Martínez, 2025). In English as a Foreign Language (EFL) contexts, the appeal of GenAI is especially strong. GenAI tools can provide immediate language explanations, grammar and writing feedback, vocabulary support, and interactive practice that may be difficult to obtain consistently in time-limited classrooms (Ekizer, 2025; Kundu & Bej, 2025; Zadorozhnyy & Lai, 2024). These benefits align with EFL learners’ needs for frequent input, practice, and feedback. At the same time, EFL learning is fundamentally social and interactional, and students may experience tension between the convenience of AI assistance and the human dimensions of language teaching (e.g., relational support, cultural mediation, and pedagogical judgment) (Almashour et al., 2025; Guan et al., 2025). In Iran, emerging evidence suggests that university EFL students view GenAI as useful for language development while simultaneously expressing concerns about accuracy, dependence, and the need for stronger evaluation skills when using AI-generated content (Rezai et al., 2024).
A key construct connecting the opportunities and risks of GenAI in learning is epistemic trust—the extent to which learners judge communicated information as
|