Scaffolding Academic Writing Using AI Assistance: A Systematic Literature Review

Abstract

In spite of the fast expansion of studies in this area, knowledge about how AI serves as an efficient scaffold for academic writing is fragmented and not systematically synthesised. To fill the gap, this paper provides an extensive systematic literature review analysing the role of AI assistance as a method to scaffold academic writing. The review of the articles was performed following the PRISMA guidelines for methodological transparency and consistency. Next, a series of advanced search strategies structured around 3 thematic keyword clusters was used to systematically search two major academic databases (Web of Science and Scopus) for studies published between 2020 and 2026. The themes found were: (1) learner engagement, agency, and affective responses in AI-assisted academic writing; (2) perceptions, acceptance, and ethics of AI-technologies in academic writing; and (3) pedagogic practices, instructional design & automated feedback in AI-supported writing. After identification, screening, eligibility, and quality assessment, 23 original studies remained for analysis. The reviewed evidence suggests that AI-assisted scaffolding could potentially contribute to academic writing development through fostering motivation, agency, and reducing affective barriers (e.g., writing-anxiety), provided that AI tools are integrated into well-designed instructional frameworks. In addition, the included studies stress a relatively positive outlook on AI technologies by learners and educators with ongoing ethical concerns of IDA tied to academic integrity, dependence, and authenticity. In addition, pedagogically informed teaching-learning environments and mindful application of automated feedback are found to be key elements in enhancing AI-supported writing. Overall, this review suggests that AI can have its greatest impact as a support structure for scaffolding rather than a substitute for instructional guidance, with implications applicable to both practice and policy in the implementation of AI in academic writing instruction going forward.