<![CDATA[While mobile-assisted language learning (MALL) and artificial intelligence (AI) are known to enhance EFL learning, there exists a deficiency in comprehending their exact influence on discrete speaking sub-skills. This study addresses this gap by providing a granular analysis of how the AI-powered application ELSA Speak affects specific aspects of speaking proficiency and motivation. Employing a quasi-experimental, pretest-posttest design over six weeks, 20 Indonesian EFL learners were divided into an experimental group (n=10) using ELSA Speak and a control group (n=10) receiving conventional instruction. The results showed a notable enhancement in the speaking abilities of the experimental group, with statistical significance (t(9) = 6.025, p < .001), whereas the control group showed no significant improvement. The mean score for the experimental group increased from 58.6 to 68.7, with fluency showing the most substantial gain (+18.3 points), followed by intonation (+16.8) and pronunciation (+8.1). Furthermore, questionnaire data indicated very high motivation (M=4.49/5.00). A significant limitation of this study is its limited sample size, which impacts the generalizability of the results. The primary practical implication is that AI tools like ELSA Speak are highly effective as supplementary resources for building learner fluency and confidence. More extensive studies with broader demographics and longer time frames are required to validate the results.]]>
