Artificial intelligence integration and self-regulated learning in English as a foreign language: Scaffold or substitute

Abstract

Artificial intelligence integration has increasingly been positioned as a mechanism for supporting self-regulated learning (SRL) among English as a Foreign Language (EFL) learners, yet the empirical evidence remains scattered across tool types and contexts. This systematic review synthesized 22 peer-reviewed studies published between 2020 and 2026, comprising 13 experimental or quasi-experimental, 6 mixed-methods, and 3 qualitative designs, examining what AI tools and frameworks characterize the literature, how AI integration affects SRL dimensions, and what factors moderate the AI-SRL relationship. Following PRISMA 2020 guidelines, studies meeting seven inclusion criteria, spanning population, design, document type, language, timeframe, and validated SRL measurement, were retrieved from Scopus and Web of Science, screened with substantial inter-rater agreement (kappa = .83 abstract stage, .79 full-text), and appraised using the Mixed Methods Appraisal Tool, whose single framework accommodates all three design types. The AI tool landscape has shifted markedly. Generative AI platforms, particularly ChatGPT and its derivatives, now account for 64% of studies, compared to 18% for earlier automated writing evaluation tools. AI integration most consistently supported goal-setting, self-monitoring, and metacognitive strategy use, while motivational regulation proved less stable, with five of 22 studies reporting reduced autonomous motivation among high-AI-use learners. AI functions as an effective regulatory scaffold within structured pedagogical sequences, though benefits depend on tool type, learner profile, and EFL context. Future research should examine motivational regulation and metacognitive cycles longitudinally, particularly in underrepresented Asian EFL settings.