A new movement is sweeping through fanfiction circles. The goal: root out writers using generative AI. The methods? Questionable at best.
On June 29th, an anonymous X account called @heatedrivalryai posted a skin for Archive of Our Own (AO3) that claims to detect Claude-generated text. The idea is simple — when text is copied straight from Anthropic’s Claude into AO3, the bot leaves behind a hidden code tag: ‘font-claude-response-body’. The skin turns the entire screen red if it finds that tag.
We tested it. It works. Paste directly from Claude? Red screen. Paste the same text through a text editor first? No red. But here’s the problem — this only catches raw, unedited pastes. Anyone who runs their work through Google Docs or Word before posting is invisible to this tool.
And the red screen doesn’t tell you how much AI was used. Could be a fully generated fic. Could be someone who ran a few sentences through Claude for spell-checking. Same result, same scarlet page.
That nuance hasn’t stopped the mob. Writers are being publicly named and shamed. Some have updated their works to remove the artifacts. Future works can easily evade detection.
The creator says the intent isn’t to create “an environment of mistrust.” But that’s exactly what’s happening. Fan communities are now policing each other with a tool that’s trivially bypassed and incapable of measuring degree of use.
Here’s the irony: AI detection tools have never been reliable for text. OpenAI shut down its own detector because it was terrible. Google’s SynthID works for images and audio, not copy-pasted prose. Even the article’s author notes they’ve been reporting on this problem for years and no universal solution exists.
AO3 already has a tag: ‘Created Using Generative AI’. Some authors use it. Most don’t — because there’s no incentive to self-report when the backlash is brutal.
One commenter pointed out the obvious: AI writes like that because it was trained on stuff real people wrote. People have been writing overly flowery fanfic since long before ChatGPT existed. Vibes-based detection will catch plenty of innocent writers.
Fandom is supposed to be a hobby, not a surveillance state.
