AI-Generated Feedback in Secondary EFL Writing: A Scoping Review of an Emerging Evidence Base

Abstract

Generative artificial intelligence (GenAI) has become increasingly prominent in second language writing instruction, yet empirical research on AI-generated written corrective feedback among secondary-school English as a Foreign Language (EFL) and English as a Second Language (ESL) learners remains limited. This scoping review mapped the available empirical evidence on AI-generated feedback and AI-generated text in secondary and middle-school EFL/ESL writing. The review was guided by the Joanna Briggs Institute scoping review framework and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR). The review employed a Population–Concept–Context framework, focusing on secondary and middle-school EFL/ESL learners, AI-generated written corrective feedback and related automated writing support, and formal secondary or middle-school writing instruction. Searches were conducted iteratively in June 2026 using a general academic web-search tool, and potentially relevant sources were screened and verified against their original publications. A single reviewer conducted the search, screening, and verification. Two studies met the eligibility and verification criteria: Ekizoğlu and Demir (2025), involving 60 Turkish secondary-school students, and Woo et al. (2025), involving 59 Hong Kong secondary-school students across seven schools. No formal quality appraisal or risk-of-bias assessment was conducted. Data were charted according to country and educational context, sample characteristics, research design, AI intervention or tool, comparison condition, outcome measures, and principal findings, and synthesized descriptively in relation to three research questions concerning writing accuracy and revision, learner engagement, and methodological and contextual gaps. The findings showed that AI-generated feedback was associated with significantly greater gains in overall writing performance, particularly grammar, vocabulary, and coherence, in the Turkish study, while 87% of students in the experimental group reported increased confidence in their writing. In the Hong Kong study, composition length was the strongest predictor of writing quality, while the relationship between AI-generated text use and writing performance varied according to academic banding. Among lower-banded students, attempts to revise AI-generated text were sometimes associated with lower composition scores. Overall, the evidence indicates that the benefits of AI-generated feedback at the secondary level are uneven and appear to vary according to learner proficiency, academic context, and patterns of engagement with AI-generated material. The review identifies a small and methodologically heterogeneous evidence base concentrated in only two national contexts and highlights the need for further research to establish the conditions under which AI-generated feedback may support secondary-school EFL/ESL writing development.