Before K-pop fanfiction tells stories about idols, it teaches fandom how to read them. This article examines that work as narrative labor, the collective transformation of public idol materials into shared conventions of character, relation, and affect. Rather than treating fanfiction as a derivative extension of official content, it asks how fans make idol personae legible through names, gestures, bodily signs, pairings, and recurrent scenes. Drawing on official and paratextual materials and a curated corpus of approximately 1,350 K-pop fanfiction texts, the study combines computational text analysis with interpretive reading. The analysis shows that fanfiction does not simply reflect idol images. It reorganizes them into relational scripts and affective archives that structure recognition, intimacy, hierarchy, desire, and memory. By reframing K-pop fanfiction as an affective infrastructure of legibility, the article contributes to K-pop studies, fanfiction studies, and digital humanities. It also offers an account of fan narrative labor before generative AI became a normalized condition of cultural production, clarifying what platform and AI systems encounter, simulate, or reorganize.