AI dependency and its impact on EFL students’ authentic language performance: A qualitative exploratory study

The widespread adoption of AI tools such as ChatGPT, Gemini, and DeepSeek by university students has raised pressing questions about their implications for authentic language development in EFL contexts. This qualitative exploratory study investigates how EFL students use AI tools in their language learning tasks, how AI dependency affects their authentic language performance, and what awareness they have regarding responsible AI use. Data were collected from 136 EFL students across three discipline-specific courses at Universitas Siliwangi, Indonesia, through open-ended questionnaires supplemented by semi-structured interviews with five purposively selected participants. Reflexive Thematic Analysis (Braun & Clarke, 2006; Braun & Clarke, 2022) yielded six themes: AI as a Cognitive Scaffold; Surface-Level AI Use; Authentic Performance Anxiety; Cognitive Atrophy; The Knowing-Doing Gap; and Pedagogical Void. The findings revealed that while most students described using AI as a starting point rather than a final product, a significant subset engaged passively with AI-generated content, resulting in diminished confidence and impaired spontaneous language performance. Furthermore, 77.2% of the participants acknowledged that AI dependency weakened their capacity to think and express ideas in English independently. Students demonstrated conceptual awareness of responsible AI use but reported consistent difficulty in translating this awareness into practice, identifying the absence of structured pedagogical guidance as the primary obstacle. These findings contribute to the growing literature on AI in EFL education and have concrete implications for AI literacy instruction and authentic assessment design in Indonesian higher education.