Year: 2026 | Month: August | Volume: 13 | Issue: 8 | Pages: 40-50
DOI: https://doi.org/10.52403/ijrr.20260804
Students' Acceptance of an AI-Powered Chatbot for Pronunciation and Speaking Practice in a Rural Indonesian EFL Classroom: A TAM-Informed Mixed-Methods Study
Dian Sari Marchia1, Sri Wahyuni2
1,2Master's Program of English Language Education, Faculty of Languages and Arts, Universitas Negeri Semarang, Semarang, Indonesia.
Corresponding Author: Dian Sari Marchia
ABSTRACT
This study assessed students' adoption of an AI-powered chatbot deployed to improve pronunciation and English speaking practice in a rural Indonesian senior high school. Perceived utility, perceived ease of use, and desire to continue using the technology were interpreted using a TAM-informed embedded mixed-methods methodology. Nineteen 10th-grade students engaged in nine educational sessions of voice interaction with ChatGPT, speech shadowing, and a genre-based approach. Quantitative data were collected using a 15-item Likert scale questionnaire and pre-test-post-test speaking tests, while qualitative data were collected using semi-structured interviews. Overall, the questionnaire was evaluated quite positively (M = 4.27/5; Cronbach’s alpha =.838). Vocabulary development obtained the highest rating (M = 4.44), followed by feedback effectiveness (M = 4.37), fluency (M = 4.26), overall effectiveness (M = 4.18), and confidence (M = 4.09). Speaking scores also improved considerably from M = 28.87 to M = 64.54, t(18) = 9.874, p <.001, Cohen's d = 2.265. Interviews revealed that perceived utility was high due to pronunciation modeling, vocabulary increase, instant feedback, customization, and repeated practice. The more I used it, the easier it got, but there were still some difficulties around building prompts, understanding native speaker accents, internet connectivity, and sometimes receiving irrelevant results. Fifteen students chose AI-assisted learning, and fourteen reported greater AI practice post-intervention, showing good behavioral intention. The findings suggest that usefulness was the strongest basis of acceptance, while ease of use was conditional on scaffolding, infrastructure, and teacher mediation. AI chatbots can therefore complement, rather than replace, teachers in rural EFL speaking instruction.
Keywords: AI-powered chatbot; Technology Acceptance Model; pronunciation; speaking skills; EFL; rural education.
[PDF Full Text]