Fanfiction readers are building ways to hunt for AI inside stories that were supposed to belong to the intimacy of fandom. The Verge reports that an anonymous X account released an AO3 skin that turns a work page red when it detects a Claude-related code artifact pasted directly into Archive of Our Own. The method can catch one specific trace, and The Verge's own testing found that the red screen appeared when Claude text was copied straight into AO3. It can also miss generated work that passes through another editor, and it cannot tell whether Claude wrote an entire story, corrected a paragraph, translated a line, or touched a draft through someone else's editing.

The tool changes the reading posture. A fanfic community built from pseudonyms, gifts, borrowed characters, private obsession, and volunteer labor starts treating style as evidence. Readers inspect punctuation, sentence rhythm, purple prose, and hidden code, then some name and shame writers in public. The fear is understandable. Generative AI feeds on the same open web where fanworks live, and many writers see undisclosed use as a violation of the human exchange that keeps fandom alive. The damage comes when suspicion becomes the main interface. A strange sentence stops being clumsy, excessive, young, translated, or simply personal. It becomes a possible confession.