With the growth of chatbots, concerns about implementing artificial intelligence (AI) chatbots in educational settings have consistently arisen, especially for the purpose of language learning. This study introduced a task-based voice chatbot called “Ellie”, newly developed by the researchers, and examined the appropriateness of its task design and performance as an English conversation partner and students’ perceptions on using it in EFL class. Korean EFL learners (N = 314) aged 10–15 years performed three speaking tasks with Ellie in their school classroom. The participants took 9.63 turns per session on average using the first 1,000-word band, indicating that the chatbot highly encouraged students to engage in conversation, which rarely occurs in general EFL classes in Korea. The high task success rates (88.3%) showed the design appropriateness of both L2 tasks and operational intents in terms of users’ successful understanding and completeness of the given chatbot tasks. The participants’ responses to the survey not only supported the positive potential of the chatbot in EFL settings but also revealed limitations to be resolved. Future suggestions for advancing and implementing AI chatbots in EFL classrooms are discussed.

随着聊天机器人的发展，人们对在教育环境中实施人工智能 (AI) 聊天机器人的担忧不断出现，尤其是出于语言学习的目的。本研究介绍了研究人员新开发的基于任务的语音聊天机器人“Ellie”，并检验了其作为英语会话伙伴的任务设计和性能的适当性以及学生在英语课堂上使用它的看法。韩国外语学习者（N= 314) 10-15 岁的孩子在学校课堂上与 Ellie 一起完成了三项口语任务。参与者使用前 1,000 字带平均每节课轮换 9.63 次，这表明聊天机器人高度鼓励学生进行对话，这在韩国的普通 EFL 课程中很少发生。高任务成功率 (88.3%) 显示了 L2 任务和操作意图在用户对给定聊天机器人任务的成功理解和完整性方面的设计适当性。参与者对调查的回应不仅支持了聊天机器人在 EFL 设置中的积极潜力，而且还揭示了有待解决的局限性。讨论了在 EFL 课堂​​中推进和实施 AI 聊天机器人的未来建议。