<![CDATA[This descriptive qualitative study examined how five Indonesian high school students (S1–S5) used Gemini AI as a tool for preparing Analytical Exposition speeches in English. A theme coding scheme comprising seven interaction categories: Generate (GEN), Revise (REV), Explain (EXP), Simplify (SIM), Practice (PRAC), Pronunciation (PRON), and Vocabulary (VOC) was used to analyze the data obtained from verbatim chat conversation logs. The results showed that GEN was the most common pattern (11 out of 35 total questions), indicating students' tendency to assign content creation to AI. On the other hand, student autonomy and metacognitive awareness were stronger among students who participated in REV and PRAC-oriented exchanges. The most beneficial interaction pattern in terms of education was S1's usage of AI as a conversational speaking partner. S5's revision-dominated interactions showed the strongest writing autonomy. The study comes to the conclusion that in order to move students from passive content consumption to active, continuous AI engagement, pedagogical scaffolding is crucial.]]>
Copyrights © 2026
