<![CDATA[The integration of artificial intelligence (AI) platforms into English for Specific Purposes (ESP) instruction has opened new avenues for sustaining oral practice beyond classroom hours. However, ESP learners in Indonesian higher education, particularly in non-English departments such as sport coaching, frequently exhibit low daily speaking habits, limited exposure to domain-specific discourse, and high anxiety in face-to-face oral practice. This study investigates the effectiveness of AI-mediated micro-speaking tasks (short, structured, daily oral activities of 60–180 seconds) in enhancing speaking proficiency, speaking habit formation, and learner self-efficacy among ESP students. Employing a quasi-experimental mixed-methods design, the research involved 64 second-year students of the Department of Coaching, Universitas Negeri Padang, divided into an experimental group (n = 32) using AI platforms (ChatGPT Voice, ELSA Speak, and Google Gemini Live) and a control group (n = 32) receiving conventional speaking instruction. Data were collected through pre- and post-tests of speaking proficiency, a daily-habit log, the Speaking Self-Efficacy Scale, and semi-structured interviews. Quantitative analysis using paired and independent samples t-tests revealed a statistically significant difference between groups (t = 6.42, p < .001, Cohen's d = 1.61), with the experimental group gaining 18.7 points compared to 7.4 points in the control group. Qualitative findings indicated reduced anxiety, increased autonomy, and stronger habit consolidation. The study contributes a replicable micro-task framework for ESP contexts and demonstrates that brief, AI-scaffolded daily practice can transform speaking from an episodic classroom task into a sustained personal habit.]]>
Copyrights © 2026
