<![CDATA[The integration of Artificial Intelligence (AI) into English language learning marks a significant pedagogical shift, yet its implications for secondary-level learners in resource-constrained contexts remain underexplored. This study examines how secondary students in semi-urban Nepal perceive AI tools—ChatGPT, Google Translate, Deep Seek, and Gemini—through the lens of Vygotsky's Zone of Proximal Development (ZPD), asking whether AI functions as a scaffold that enhances linguistic competence or a crutch that erodes autonomy. Using a phenomenological design, data were collected from twenty-three purposively selected students through semi-structured interviews, analyzed via Braun and Clarke's thematic analysis. Findings reveal a dual role: AI extended students' capacities within the ZPD through instant feedback, corrective support, and confidence-building, yet simultaneously fostered dependency, superficial engagement, and academic dishonesty when used without guidance. Students navigated this tension by critically evaluating AI outputs, adapting them to local context, and relying on teacher mediation to ensure ethical use. Persistent infrastructural inequality further constrained equitable engagement across participants. The study underscores teachers' indispensable mediating role in transforming AI from a shortcut into a genuine scaffold, and calls for structured AI literacy, teacher training, and equity-oriented policy to support responsible integration in under-resourced English language classrooms.]]>
