The Role of AI-Powered Feedback in Developing Efl Learners' Speaking Accuracy

Artificial intelligence (AI) has expanded the possibilities for providing immediate, individualized and repeatable feedback to learners of English as a foreign language (EFL). This paper examines the role of AI-powered feedback in developing speaking accuracy, understood as accurate and intelligible pronunciation, grammatical control, appropriate lexical selection and effective self-correction during oral production. The analysis synthesizes evidence from international research on automatic speech recognition (ASR), AI speech-evaluation systems, speech-enabled corrective feedback and generative AI. The literature indicates that AI can strengthen the feedback cycle by making errors more noticeable, shortening the time between performance and correction, increasing opportunities for deliberate repetition, and supporting learner autonomy. Meta-analytic evidence shows a moderate positive effect of ASR on L2 pronunciation, while recent studies report gains in speaking performance and favorable learner perceptions. At the same time, the pedagogical value of AI feedback depends on recognition accuracy, the quality and specificity of explanations, learner proficiency, task design and teacher mediation. A hybrid model in which AI handles high-frequency diagnostic feedback and teachers validate, prioritize and contextualize corrections is therefore proposed as the most educationally sound approach.