| Abstract: |
Speaking anxiety continues to challenge many college ESL students, affecting confidence, fluency, and classroom participation. With the growing integration of artificial intelligence (AI) and large language models (LLMs) in higher education, new opportunities have emerged to enhance oral language learning through adaptive and low-pressure environments. This review followed the SANRA guidelines to ensure clarity, rigor, and balance. Relevant peer-reviewed studies were identified from ERIC, EBSCOhost, JSTOR, IEEE Xplore, and Scopus using the keywords artificial intelligence, large language models, ESL, EFL, speaking anxiety, and communication confidence. Research focusing on higher education contexts was analyzed for methodological quality and pedagogical insights. Findings show that AI-based tools such as chatbots, speech recognition systems, and feedback applications improve learners' fluency, pronunciation, and vocabulary while enhancing grammatical accuracy and discourse coherence. These technologies also reduce anxiety, boost motivation, and promote autonomous learning by providing immediate, personalized feedback in supportive settings. However, limitations in emotional responsiveness, contextual adaptability, and the need for teacher mediation highlight the importance of balanced integration. Overall, AI serves as a valuable complement to traditional instruction, fostering both linguistic competence and emotional well-being among college ESL learners. [For the complete proceedings, see ED678731.] |