Heo Seona This study examines the challenges encountered in Korean language classrooms employing generative AI-framing them as issues related to language-based judgment rather than technical proficiency. Based on a literature review and focus group interviews with seven Korean language teachers-the results indicated that students tend to treat AI-generated texts as completed answers, with limited recontextualization, verification, responsibility attribution, and interaction. Accordingly, the study proposes four components of Korean language-based AI literacy: recontextualization, verification, responsibility attribution, and meaning negotiation. Therefore, this contributes to a subject-specific AI literacy framework that emphasizes judgment and responsibility in Korean language education.