The rapid integration of artificial intelligence tools particularly ChatGPT into educational environments has introduced new demands on students' capacity to use AI critically, ethically, and autonomously. This study explored the AI literacy levels of tenth-grade students at SMAN 2 Bengkalis and examined how ChatGPT use differed between students with high and low levels of AI literacy in English language learning. An explanatory sequential mixed-method design was employed, combining a 20-item Likert-scale questionnaire administered to 96 students and semi-structured interviews conducted with six purposively selected participants the three highest and three lowest AI literacy scorers. The questionnaire was adapted from the ABCE framework (Ng et al., 2023), encompassing Affective, Behavioral, Cognitive, and Ethical dimensions. The findings revealed that students demonstrated a high overall AI literacy level (M = 2.77), with the Ethical dimension scoring highest (M = 3.39) and the Cognitive dimension lowest (M = 2.43). Interview analysis showed that high-literacy students used ChatGPT with intentional learning goals, critical post-response evaluation, context-rich prompting strategies, and autonomous engagement, while low-literacy students predominantly relied on ChatGPT for task completion, copied responses without reflection, and reported dependency. These differences confirm that AI literacy level meaningfully moderates the educational quality of ChatGPT interactions. The study concludes by recommending the integration of explicit AI literacy instruction, prompt engineering pedagogy, and AI ethics discussion into English language classrooms.