AI tools in academic writing: a PRISMA-based systematic review in applied linguistics

Generative artificial intelligence (AI) has prompted higher education institutions to reconsider governance, pedagogy, and ethics surrounding AI-based writing tools. This PRISMA-guided systematic review synthesises empirical research on AI-assisted academic writing in language education and multilingual higher-education contexts. It addresses three questions: which AI tools are used, what benefits and challenges are reported, and what ethical, authorship, and transparency issues arise? The review covers empirical studies indexed in Scopus, Web of Science, IEEE Xplore, and ERIC between 2020 and 2025 (N = 37) and applies thematic synthesis informed by socio-cognitive and ecological perspectives on writing. Findings indicate support for productivity, perceived linguistic quality, and reduced affective barriers among non-native English speakers, alongside risks of overreliance, cognitive offloading, hallucinated content, unequal access, blurred authorship, inconsistent disclosure, and limited institutional guidance. The review proposes a three-domain framework, productivity, cognition, and ethics, and derives implications for AI literacy, process-oriented assessment, and institutional policy.