Innovation in Language Learning

Artificial intelligence (AI) is increasingly used in foreign language education to expand oral practice, personalize feedback, and support learner autonomy. For EFL learners, speaking is among the most difficult of skills to acquire as it involves real-time integration of a number of factors such as fluency, pronunciation, grammatical accuracy, interactional competence and confidence. This paper presents a systematic literature review of AI-based pedagogical strategies used to improve speaking skills in EFL/ESL contexts from 2018-2025. A PRISMA-informed review with the PICOC framework was utilized. The final extraction matrix was limitedto no more than one copy of each reference (i.e., one per author).The final database contained more than 400 entries and included new studies in addition to replacement studies from previously discovered/deleted studies obtained from reliable academic sources, including System, Computers & Education, Computers and Education: Artificial Intelligence, Innovation in Language Learning and Teaching, Frontiers in Psychology, Computer Assisted Language Learning, Computers in Human Behavior, and British Journal of Educational Technology (the full listing of these publications can be found by accessing the corresponding reference number included with each project listed at the end of this report). Based on the inclusion/exclusion criteria, 30 results were finally synthesized based upon the inclusion criteria metrics. The most frequent technologies found in the synthesis included conversational chatbots; generative AI; automated speech recognition (ASR) systems; speech evaluation tools using AI; and adaptive/informal AI-mediated tools for learning in the classroom. The most commonly reported outcome variables found were fluency, pronunciation accuracy, motivation, autonomy in learners, willingness to communicate, confidence, and decreased levels of anxiety. AI-powered speaking strategies can enhance oral language development through pedagogically designed instructional activities and teacher mediation. However, available evidence is limited based on short duration intervention periods, a concentration of research from Africa, and the lack of longitudinal research on this area. Future research should explore speeches that enhance AI-mixed learning in higher education settings in Latin America, as well as develop responsible ethical implementation frameworks that account for cultural context differences.

Keywords: artificial intelligence; EFL; speaking skills; chatbots; automatic speech recognition; generative AI; systematic literature review