Effects of generative AI assisted grammar instruction on Ethiopian EFL learners’ self-efficacy, metalinguistic awareness, and writing performance

Despite the rapid growth of generative artificial intelligence (GenAI) in English as a Foreign Language (EFL) writing research, t​he intersection of AI-supported grammar instruction, metalinguistic reflection, and writing self-efficacy remains virtually unexplored in Ethiopian higher education contexts. This quasi-experimental study investigated the effect of structured GenAI-supported grammar instruction on the academic writing performance, metalinguistic reflection, and writing self-efficacy of first-year Natural Science EFL students (N = 88; experimental n = 45, comparison n = 43). The study employed a ten-week quasi-experimental pretest–posttest non-equivalent control group design. The experimental group engaged in a structured four-stage cycle (draft–prompt–reflect–revise) using ChatGPT, while the comparison group received conventional teacher-marked instruction. Data were collected through an analytic writing test scored across five dimensions, a metalinguistic reflection questionnaire, and a writing self-efficacy questionnaire. Multivariate analysis of covariance (MANCOVA) with pre-test scores as covariates revealed statistically significant group differences across all three outcome clusters. The experimental group outperformed the comparison group on overall writing performance (p <.001), with significant univariate effects across all five writing aspects. Significant group differences were also observed for metalinguistic reflection (p <.001) and writing self-efficacy (p <.001). These findings confirm all three research hypotheses and demonstrate that structured GenAI-supported grammar instruction produces large, practically significant benefits for Ethiopian EFL learners’ writing accuracy, conscious linguistic reflection, and academic writing confidence. The results extend the global GenAI-and-writing literature into Ethiopian EFL context and carry implications for the pedagogically scaffolded integration of AI tools in resource-constrained university writing instruction.