This year-long exploratory study examines an AI-supported personalized reading approach aimed at improving low student engagement. Conducted with 63 secondgrade high school students in a rural area, the researcher, a high school English teacher, utilized AI to generate engaging reading passages at various proficiency levels. Students were allowed to choose texts that best matched their abilities. A mixedmethods design was employed, collecting mock test scores to identify achievement trends and using comparative surveys to evaluate student perceptions compared to the previous grammar-translation method. The findings revealed a positive trend in academic achievement, marked by a decrease in low-level students and an increase in middle and high-level students. Additionally, students reported a significant improvement in their learning experience compared to the prior method, demonstrating much higher levels of satisfaction, motivation, and interest with the new approach. Notably, the autonomy students had in selecting passages successfully preserved their self-esteem, addressing a key challenge in traditional ability grouping. While acknowledging its exploratory nature and context-specific limitations, this research presents a practical model for revitalizing secondary English education and offers a rare long-term perspective on AI integration in an authentic K-12 classroom.