The integration of artificial intelligence in the academic writing of EFL students in higher education: a case study at Fergana State University

The rapid growth of AI-assisted writing tools such as ChatGPT, Grammarly and QuillBot is reshaping academic writing practices in higher education. While these technologies offer immediate linguistic support for English as a Foreign Language (EFL) learners, they also prompt ongoing pedagogical debate concerning learner autonomy, critical engagement and potential overreliance. Despite expanding global scholarship, little empirical research has examined how AI integration unfolds within post-Soviet Central Asian EFL contexts, where distinct pedagogical traditions and linguistic backgrounds may mediate its impact. This convergent mixed-methods case study explores how 351 EFL students at Fergana State University, Uzbekistan, used AI tools in academic writing and investigates the effects on writing quality, cognitive independence and technology acceptance. Drawing on Sociocultural Theory (SCT) and the Technology Acceptance Model (TAM), the study combined questionnaire data, comparative writing tasks and semi-structured interviews (n = 30). Quantitative analyses indicate that AI assistance was associated with substantial improvements in grammatical accuracy (35% error reduction, d = 1.27) and lexical sophistication (53% increase in Academic Word List coverage, d = 1.40). At the same time, reductions were observed in warranted argumentative claims (24% decline, d = 1.06) and measures of originality (d = 0.96). Perceived usefulness emerged as the strongest predictor of continued AI use (β = 0.47, p < 0.001). Qualitative findings revealed tensions between scaffolding and substitution, a perceived performance–competence gap, institutional ambiguity and increasingly strategic use among advanced learners. The findings suggest that AI may enhance surface-level performance while being associated with weaker indicators of independent academic reasoning under assisted conditions.