<![CDATA[In self-directed learning, EFL learners face persistent fluency challenges, particularly when access to native speakers or teachers is limited. Recent advancements in generative AI, such as Gemini AI, offer transformative potential by simulating human-like interactions and adapting to individual learner needs. This study investigates how Gemini AI uniquely facilitates speaking fluency through Task-Based Learning in self-directed learning and its impact on students’ speaking performance. This study involved 19 participants enrolled in an English Discussion class at the Universitas Teknologi Yogyakarta who engaged in a four-week intervention program. Utilizing a mixed-methods design, the study combined quantitative analysis of speaking fluency metrics with qualitative examination of student reflection reports. Participants completed weekly speaking tasks using Gemini AI as a conversational partner, documenting their experiences in a structured report. Quantitative results showed a 13% average improvement in speaking fluency, particularly in lexical diversity and reduced hesitation. Qualitative analysis revealed five key themes: increased confidence in spontaneous speech, appreciation for 24/7 accessible practice, effective feedback to improve students' speaking skills, enhancing vocabulary, and fostering a non-judgmental learning environment. The findings suggest that Gemini AI can effectively supplement classroom instruction for speaking skill development. However, the generalizability of findings is constrained by the small number of participants and the short intervention period. Future research should employ longitudinal designs with larger cohorts to substantiate the long-term efficacy of AI-assisted task-based language teaching. Subsequent studies should also systematically investigate strategies to mitigate Gemini shortcomings, including optimizing feedback relevance and turn-taking mechanics for educational dialogue.]]>
Copyrights © 2026
