Impacts of artificial intelligence on willingness to communicate and its key antecedents in EFL contexts: a systematic review (2015–2026)

Introduction Willingness to Communicate (WTC) is a core psychological construct in English as a Foreign Language (EFL) learning and acquisition, and Artificial Intelligence (AI) holds significant potential to enhance both learners' WTC and its key antecedents. However, existing systematic reviews have not provided a focused synthesis of AI-WTC research in EFL contexts, nor have they examined in depth how AI impacts WTC and its antecedents or offered clear directions for future research. Methods Guided by three core research questions, this PRISMA-aligned systematic review synthesizes 73 empirical studies published between 2015 and February 2026 to examine research characteristics, assess AI's impacts on WTC and its key antecedents, and identify actionable gaps for future research. Results The findings show that AI-WTC research in EFL contexts has expanded rapidly, with studies concentrated mainly in Asian EFL contexts and frequently focused on generative AI chatbots. Evidence also suggests that AI may positively influence EFL learners' WTC through both direct and indirect routes. Specifically, this review proposes an integrated framework in which AI directly enhances learners' WTC across different tool types, contexts, and age groups, while also indirectly promoting WTC through four proposed pathways: the Affective Foundation Pathway, the Motivational Propensity Pathway, the Communicative Capacity Pathway, and the Contextual Affordance Pathway. Discussion Anchored in MacIntyre et al.'s pyramid model, this framework extends WTC theory to AI-mediated EFL learning by conceptualizing AI as a dynamic, multi-layered affordance that shapes the affective, motivational, communicative, and contextual antecedents of communication. The review also identifies five interrelated gaps across methodological, contextual, population, technological, and theoretical dimensions, and highlights the need for longitudinal designs, triangulated WTC measures, diverse learner samples, comparative AI tool research, pedagogically guided AI integration, and dynamic, theory-driven integrated models to advance evidence-based AI integration aimed at enhancing EFL learners' WTC.