The review systematically analyzes AI-disruptive change in language education by considering multilingual settings, large language models, and new implications. After integrating the findings of 161 studies from 2015 to 2024, the review relies on the PRISMA framework to analyze progress and gaps regarding AI use tools, ethical considerations, and pedagogical integration. The findings indicated that AI technologies—conversational agents, speech recognition systems, and writing assistants—help in language learning by reducing learner anxiety, giving pronunciation support, and providing feedback in real time. LLMs such as Aya and LLaMA display scalable multilingual capabilities, with high variations in performance, mostly against low-resourced languages; these languages, however, are still underrepresented in datasets and evaluations. Ethical considerations around cultural biases, gender stereotyping of LLM outputs, and getting too reliant on automated tools emphasize the pressing need for appropriate governance frameworks. The pedagogical implications with mobile learning and gamification are promising but often lean towards accessibility, lacking the depth needed and thus worsening infrastructure inequalities in Africa and Eastern Europe. Other critical gaps identified include (1) little investigation into marginal languages, (2) lack of longitudinal studies on AI's educational impacts, and (3) neglecting socio-emotional issues as to whether AI may erode human interaction.
Fatimah Albedah (Fri,) studied this question.