AI-assisted tools and Pronunciation

Abstract

This descriptive study aimed to analyze learners’ perspectives regarding the use of Artificial Intelligence (AI)-assisted tools to improve English pronunciation. The research was systematically conducted with a population consisting of 121 B1-level undergraduate students enrolled in the Carrera Pedagogía de los Idiomas Nacionales y Extranjeros (PINE) at Universidad Técnica de Ambato, Ecuador. The data collection process involved the online distribution of a structured, 22-item Likert-scale survey. Content validity was established by five academic experts using Aiken’s V (0.923) for relevance, clarity, and coherence. Additionally, a pilot test was validated through Cronbach’s alpha (0.935). Empirical data analysis, focused primarily on mean scores, revealed that the most frequently used functional features are audio models and pronunciation examples in tools like YouGlish, followed closely by immediate feedback in the tool Cake. In contrast, automated speech recognition in tools like Google-Speech-to-Text was the least popular feature used due to limited interactive guidance. Regarding learning strategies, students relied on adjustment on feedback, model imitation, and repetition of pronunciation tasks, indicating learner dependence on real-time corrective guidance. On the other hand, the self-recording of pronunciation strategy received the lowest score. In terms of suprasegmental features, the most developed were intonation and rhythm. These findings suggest that while AI-assisted tools effectively promote contextual clarity and natural speech flow, structured pedagogical interventions must place more emphasis on self-directed monitoring practices and mastering syllable patterns to improve overall communicative competence.