<![CDATA[Pronunciation plays a crucial role in effective English oral communication. However, vocational high school students in Islamic boarding school (pesantren) settings often struggle with pronunciation accuracy due to native language interference, limited classroom practice time, and high speaking anxiety. Therefore, this study aimed to investigate the effectiveness of using the AI-based ELSA Speak application in improving the English pronunciation accuracy of 10th grade students at SMK Islam Terpadu Marifa Tasikmalaya. A quantitative approach with a pre-experimental one-group pretest-posttest design was employed in this study. The research involved 21 10th grade Computer and Network Engineering (TKJ) students selected through saturated sampling (N = 21). Data were collected using oral pronunciation tests administered before and after a five-meeting intervention, assessing both segmental features (vowels and consonants) and suprasegmental features (word stress and intonation). Content validity of the instrument was verified through Aiken’s V, and reliability of the instrument was verified through Cronbach Alpha and Cohen Kappa (Weighted Kappa). The data were analyzed using descriptive statistics, the Shapiro-Wilk normality test, a Paired Sample T-test, and Normalized Gain (N-Gain) analysis. The research results revealed a statistically significant increase in the mean pronunciation accuracy score from 57 on the pretest to 73 on the post-test (p < 0,001). The paired sample t-test demonstrated t(20) = -12,573 p < 0,001, leading to rejection of the null hypothesis (H0) and acceptance of the alternative hypothesis (Ha). Furthermore, the N-Gain score averaged 0,4047 (40,47%), indicating a moderate level of instructional effectiveness. These findings confirm that integrating ELSA Speak into structured classroom activities provides an effective AI-assisted learning environment that enhances phonetic awareness, reduces native language interference, and lowers foreign language anxiety.]]>
