Examining the Impact of ChatGPT-Assisted Academic Writing on the Writing Performance and Critical Thinking Skills of Undergraduate Students in Pakistan

Abstract

Generative artificial intelligence (AI) has rapidly emerged in higher education, and its potential for education and teaching, particularly in academic writing and critical thinking, remains a subject of controversy. This Quasi-experimental study, using a mixed-methods approach, investigated the efficacy of utilising ChatGPT for academic writing on students' academic writing performance and whether the guided use of ChatGPT had a positive impact on students' critical thinking.  An experimental group received assignments involving brainstorming, outlining, reading suggestions, revising, and source checking using ChatGPT for 10 weeks, while the control group received the same writing tasks using conventional writing instruction. The hypothetical sample consisted of 120 undergraduates from three universities in Pakistan, with equal numbers of females and males, students from three academic years, and from different disciplines. Data were gathered using the writing outcome instrument, critical thinking assessment, semi-structured interviews with students, student reflections, and open-ended responses to the student surveys. Higher post-test achievement was observed in the experimental group across organisation, coherence, grammar, and argumentation. As students were asked to question and verify ChatGPT outputs and revise them rather than copy, gains in critical thinking were greatest in these areas of analysis and reflection. Qualitative themes identified were an increase in confidence, more coherent writing, learning facilitated by AI, reliance issues and the need for instructors' guidance. The study suggests that embedding ChatGPT within explicit AI literacy, academic integrity, and critical evaluation practices can support the development of writing in Pakistani universities.