AI-Mediated Cognitive Load in L2 Learning: A Systematic Review and Gap Analysis with Focus on IELTS Writing & 1 Speaking in India

Abstract

Artificial intelligence (AI) and digital tools are rapidly reshaping English language teaching (ELT), yet the mechanisms by which these technologies influence learners’ cognitive processes—particularly in high-stakes tests such as the International English Language Testing System (IELTS)—remain underexplored. This systematic review synthesizes peer-reviewed empirical and review literature (2015–2025) that examines AI/digital tools in L2 instruction with attention to cognitive constructs (cognitive load, working memory, metacognition) and practical outcomes for writing and speaking. Using a PRISMA-informed search strategy across Scopus, Web of Science, ERIC, PubMed/PMC, and Google Scholar, the review maps tool types (LLMs, automated writing evaluators, chatbots, speech-recognition systems), study designs (qualitative, quantitative, mixed), and cognitive measures (NASATLX, Paas scale, working memory tasks, MAI). Results show a robust and growing literature on AI in ELT but reveal four gaps: (1) scarce India-specific empirical studies linked to IELTS outcomes, (2) limited explicit use of validated cognitiveload instruments, (3) few mixed-methods or controlled experimental designs that combine task-based pedagogy with AI, and (4) little research that evaluates free/ low-cost tools in low-resource educational settings. Building on the synthesis, I propose an AI-Mediated Cognitive Load Model (AI-MCL) that explains how AI can reduce extraneous load, scaffold intrinsic load, and support germane processing during task-based IELTS practice. The paper concludes with a research agenda and methodological recommendations for researchers and practitioners seeking to evaluate AI interventions in Indian higher education and test-preparation contexts.