Generative AI in EFL Instruction: A Systematic Review of Tools, Outcomes and Evidence (2023–2026)

A synthesis is needed of generative artificial intelligence (GenAI) in English as a foreign language (EFL) instruction to distinguish reported pedagogical benefits from directly assessed learning outcomes. This systematic review has examined GenAI tool typologies, reported linguistic, pedagogical and affective outcomes, and the evidence supporting them. Following the PRISMA framework, searches were conducted in Scopus and Web of Science (WOS) and updated in July 2026. The review included 36 open-access empirical journal articles published in English between 2023 and July 2026, which examined one or more GenAI tools in EFL or closely related ESL contexts. Studies were coded by tool typology, pedagogical function, user group, outcome dimension, educational level and evidence type, and synthesised narratively. ChatGPT was the dominant tool, used in 32 of the 36 studies included (88.9%). GenAI was used primarily to support writing, grammar, speaking and pronunciation, autonomous learning, teacher material development, assessment, creativity-oriented activities and culturally responsive curriculum design. Reported outcomes were generally positive but varied according to the supporting evidence. Some studies assessed writing and grammar performance directly, whereas reported benefits for motivation and autonomy were primarily supported by perceived or self-reported evidence. Reported concerns included overreliance, plagiarism and inaccurate information. The synthesis suggests potential benefits from pedagogically guided GenAI use, while highlighting the need to assess learning of perceived usefulness independently. Implications include critical evaluation of AI feedback and assessment of the revision decisions of learners. Future research should examine sustained learning outcomes, critical thinking and dependence through longitudinal and comparative designs.