Effects of dialogic and monologic AI/non-AI feedback on L2 speaking development: Exploring learner achievements, feedback literacy, and interactional moves

Although prior studies have examined the role of artificial intelligence (AI) in supporting second language (L2) speaking development, limited attention has been paid to how feedback modes (i.e., dialogic vs. monologic) and delivery media (i.e., AI vs. non-AI) differentially shape learners’ oral performance and feedback literacy. Addressing this gap, the present study investigated a total of 120 intermediate learners of English who were recruited from language institutes and randomly assigned to four equal groups: AI-dialogic (ChatGPT), AI-monologic (ELSA Speak), non-AI-dialogic (Zoom), and non-AI-monologic (Audacity). Data were collected through a pretest–posttest speaking assessment, a feedback literacy questionnaire, semi-structured interviews, and analysis of recorded dialogic interactions. Quantitative findings revealed that the AI-dialogic group outperformed the other groups in both speaking proficiency and feedback literacy, followed by the non-AI-dialogic group, whereas the monologic groups made comparatively weaker gains. Qualitative analysis revealed that dialogic conditions facilitated clarification requests, confirmation checks, and self-repair, whereas monologic conditions hindered learner uptake and interaction. These findings demonstrate that the mode of feedback and its medium of delivery play a crucial role in oral development, carrying both theoretical and pedagogical implications for designing hybrid feedback practices that integrate AI’s precision with the co-constructive power of dialogue.