Answers are grounded in this paper and cite their sources.
Students who write in English as a foreign language (EFL) use generative AI (GenAI) heavily, often where institutional guidance is limited. We know little about how they decide what is acceptable help and what is misconduct. This convergent mixed-methods study asked Pakistani EFL university students about academic integrity, their AI-assisted writing practices, their awareness of guidance, and the benefits and concerns they see. Sixty-seven students (48 female; 64.2% MPhil) were recruited by convenience sampling and completed an online questionnaire between May and September 2026. Likert responses were analyzed (N = 67) with descriptive statistics, Wilson confidence intervals, paired Wilcoxon tests and Spearman correlations (Holm-adjusted). Open-ended responses analyzed (n = 54 analyzed) thematically. Most respondents accepted GenAI for grammar correction (85.1%) and brainstorming (73.1%), and most rejected submitting AI-generated content without revision (68.7% disagreed). However, 85.1% said they always revise AI output, but only 41.8% said they clearly mention using it. At the same time, 70.1% saw unacknowledged use as similar to plagiarism (paired difference p < .001). Just over half (52.2%) agreed that their university gives clear guidelines. In contrast, 71.6% said they understand ethical use and 68.7% wanted regulation. Concern about incorrect information (79.1%) and harm to learning (76.1%) was common, but few admitted dependence (29.9%). Seven themes described supportive but limited use, gains in efficiency and understanding, heavier delegation by a minority, the burden of checking output, fear of losing skills, practical barriers, and calls for rules and AI-literacy teaching. Five of 53 respondents wanted restriction. The results suggest a gap in disclosure and guidance more than a problem of compliance. The sample was self-selected and the data are self-reported, so the findings are exploratory.
No takes yet. Share an insight, caveat, or question.
Shahid et al. (2026) studied this question.