Exploring EFL learners’ perceived promise and limitations of using an artificial intelligence speech evaluation system for speaking practice Bin Zou a , Sara Liviero b , Qing Ma c , Weilei Zhang d , Yiran Du e,* , Peiling Xing f a Xi’an Jiaotong-Liverpool University, Department of Applied Linguistics, Suzhou, Jiangsu, China b Durham University, Durham, UK c The Education University of Hong Kong, Hong Kong, China d Shanghai University of International Business and Economics, School of foreign languages, Shanghai, China e Harvard University, Cambridge, MA, USA f Xi’an Jiaotong-Liverpool University, English Language Centre, Suzhou, China A R T I C L E INFO Keywords: Artificial intelligence (AI) Speech evaluation system Speech scoring English as a foreign language learning (EFL) Speaking skills Learner perceptions ABSTRACT This study explores English as a Foreign Language (EFL) learners’ perceptions of the promise and limitations of EAP Talk, an AI-based speech evaluation system, for speaking practice. Using a mixed-methods approach, data were collected from 366 EFL learners across five universities through questionnaires and semi-structured interviews. The findings reveal that EAP Talk significantly enhances speaking skills, including pronunciation, grammar accuracy, idea- organization, read-aloud, and presentation skills. Participants appreciated the convenience, motivational aspects, and institutional efficiency of EAP Talk, which allowed for flexible and autonomous learning. However, several limitations were identified. Issues with the accuracy of voice recognition and score feedback affected learners’ confidence in the feedback. The relevance of practice materials to academic courses was another concern, as was the limited detail in the feedback provided by EAP Talk, which often lacked comprehensive explanations and corrective guidance. The absence of feedback on grammar accuracy and speech organization further limited the tool’s effectiveness for developing comprehensive speaking proficiency. These findings sug- gest that while EAP Talk holds significant potential for enhancing language learning, ongoing improvements are necessary to address its limitations. This study contributes to the under- standing of how AI tools can support EFL learners for speaking practice. 1. Introduction Artificial Intelligence (AI) acts as a virtual assistant tool by simplifying and enhancing everyday human activities (Russell & Norvig, 2016). Its integration into various sectors, including education, has garnered significant interest among stakeholders, leading to a surge in AI patents, particularly in speech evaluation systems (Zou et al., 2023). AI technologies featuring voice recognition and speech evaluation systems have been developed to assist English as a Foreign Language (EFL) learners in improving their speaking skills * Corresponding author. 1260 Boylston St, APT 414, MA, 02215, USA. E-mail addresses: bin.zou@xjtlu.edu.cn (B. Zou), sara.liviero@yahoo.co.uk (S. Liviero), maqing@eduhk.hk (Q. Ma), leisabrina@126.com (W. Zhang), yid164@harvard.edu (Y. Du), Peiling.Xing@xjtlu.edu.cn (P. Xing). Contents lists available at ScienceDirect System journal homepage: www.elsevier.com/locate/system https://doi.org/10.1016/j.system.2024.103497 Received 21 March 2024; Received in revised form 18 September 2024; Accepted 20 September 2024 System 126 (2024) 103497 0346-251X/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.