Ai As Support Or Substitute? Instructional And Generative Ai Use In English Academic Essay Writing Among Mainland Chinese Undergraduates At A University In Malaysia
Keywords:
artificial intelligence, English academic writing, instructional AI use, generative AI use, writing self-efficacy, higher education, Undergraduate EFL studentsAbstract
The rise of Artificial Intelligence (AI) in academic writing has sparked debates about its impact on students' confidence in their ability to write independently. As AI becomes more prevalent in academic writing, there has been an increase in discussion regarding how this technology affects the confidence students have in their writing abilities. This paper explored correlations of the self-efficacy of independent English academic essay writing and the use of instructional and generative AI technologies by Mainland Chinese undergraduates in a private international university in Malaysia. In this study, AI use was categorized as instructional use (e.g., seeking explanations, feedback, brainstorming support, writing guidance) and generative use (e.g., asking AI to write sentences, paragraphs, or other written text. Self-efficacy was investigated in the pre-writing, writing, and post-writing phase. A cross-sectional quantitative survey design with convenience sampling method was used. A total of 254 online questionnaires were distributed, of which 145 valid responses were remained for analysis. Findings revealed that the use of instructional AI was moderately high (M = 3.42, SD = 0.68), while the use of generative AI was moderate (M = 3.05, SD = 0.81). The overall writing self-efficacy was also moderate (M = 3.05) and the highest level of self-efficacy was reported in pre-writing and the lowest in post-writing. The regression analysis results revealed that the use of instructional AI positively predicted writing self-efficacy (β = .32, p < .001) and negatively predicted writing self-efficacy (β = −.21, p = .002) for generative AI use. The interaction effects between the two types of AI use were not significant, and the overall model accounted for about 24% of the variance in writing self-efficacy. The results indicate that while using instructional AI can enhance students' self-efficacy through feedback and scaffolding, relying on AI-generated content may be associated with a decrease in opportunities for students to develop their own writing skills.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Dr. Uranus Saadat (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.