Vocational high school students in Automotive Engineering increasingly need oral English for workplace communication, yet conventional classrooms offer limited opportunities for individual speaking practice. Voice-enabled generative artificial intelligence may provide supplementary, low-pressure practice, but evidence on objectively assessed speaking outcomes in vocational English for Specific Purposes (ESP) contexts remains scarce. This study examined changes in the speaking fluency and pronunciation of Grade X Automotive Engineering students after using ChatGPT Voice Mode. A quantitative one-group pretest-posttest pre-experimental design was applied to 36 purposively selected students at SMKN 1 Blora, Central Java, Indonesia. Over four weeks, students completed weekly 45-minute sessions of structured, vocationally contextualized roleplays, such as customer-mechanic consultations and troubleshooting conversations. Recorded performances were scored by two raters using an adapted analytic rubric, with good to excellent inter-rater reliability. Wilcoxon signed-rank tests were used for fluency and pronunciation, and a paired-samples t-test for the overall score. Results showed significant improvements in fluency (mean difference = 22.36), pronunciation (25.42), and overall speaking performance (23.89), all with p < .001 and large effect sizes; every student scored higher at post-test. Nevertheless, without a control group, the findings indicate change associated with the intervention rather than proof of causation. The study suggests that ChatGPT Voice Mode can serve as a supplementary speaking practice tool in vocational ESP instruction, complementing teacher-led feedback.
Copyrights © 2026