The fanfiction community is at war with AI — and itself
SkimNews Take
Detection-by-fingerprinting breaks the moment text is lightly edited, so moderation workload shifts from catching cheaters to resolving false positives — a burden volunteer communities aren't structured to absorb.
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- @heatedrivalryai released the AO3 skin on June 29th, flagging pages with red backgrounds when they contain the "font-claude-response-body" code that Claude wraps around copied text
- The Verge's testing confirmed the detector works on direct pastes from Claude, but the red screen vanishes if the same text is moved through Google Docs, Microsoft Word, or any intermediary editor
- Fanfic communities mobilized to publicly name and shame flagged writers, despite @heatedrivalryai saying the demonstrations were meant to prove functionality, not accuse specific users
- The skin can't gauge how heavily Claude was used — fully AI-generated stories and authors who pasted a few sentences through Claude for spell-checking or translation trigger the same scarlet page
- At least one writer was wrongly caught up because another person they trusted to edit their fic did so using Claude, exposing anyone in an author's toolchain to false accusation
- No reliable universal AI-text detector currently exists — invisible-watermark tools like C2PA and SynthID cover images and audio but don't carry over to copy-pasted text, and AI companies have limited public incentive to build one beyond avoiding model collapse
Why it matters: Fandom's detector only catches Claude text pasted straight into AO3, so a writer who ran a sentence through spell-check gets the same scarlet flag as one who AI-generated an entire story. AO3's voluntary "Created Using Generative AI" tag depends on author honesty, and the public-shaming backlash now incentivizes concealment over disclosure.
Ask SkimNews
