AI writing assistants and their effect on EFL writing quality.

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

In this systematic review, the impact of the AI writing assistants on the quality of writing in English as a Foreign Language (EFL) is explored, focusing on measurable outcomes and learners' experience. The review collates data from 20 recent studies conducted between 2020 and 2026 that have examined the use of Grammarly, ChatGPT, ChatGPT-4, Gemini, Copilot, and Automated Writing Evaluation Systems. The review is based on the writing elements of PRISMA 2020 which are writing accuracy, grammar, vocabulary, coherence, cohesion, organisation, argumentation, revision behaviour, confidence and teacher mediation. The results indicate that AI writing assistants are most consistent in enhancing lower-order writing skills, particularly in the areas of grammar, mechanics, spelling, punctuation, vocabulary selection, and sentence-level clarity. Evidence for higher-order writing improvement is more conditional, generative AI can help with coherence, organisation and idea development, but learners need to be reflective when using feedback, and teachers need to guide learners through the revision process. The review also reveals methodological gaps in the literature, such as the use of small samples, short intervention periods, lack of longitudinal studies, over-representation of perception data, and inconsistencies in reporting writing assessment procedures. The findings are related to the theories of Computer-Assisted Language Learning, sociocultural, cognitive load, constructivism and self-determination theory. It posits that AI-assisted writing platforms are most useful as scaffolding tools for process-based writing instruction, not as a replacement for teacher feedback nor as a stand-alone method of writing. It is concluded that using AI in EFL writing can be guided, ethical and transparent to increase EFL writing development, and future research is needed to investigate long-term transfer, unaided writing performance and comparative feedback models. It also suggests more transparent classroom policies to help students recognise when feedback is provided via AI and when it is not.

Links and sources

Comments