Why online fan writers are hunting for AI code

People have long used clues like strange word choices or repetitive sentence patterns to guess if a story was written by a computer model — a program trained on massive amounts of text to predict and generate human-like sentences. Recently, however, a new tool appeared for the site Archive of Our Own. It functions like a digital filter that scans for hidden background data called code artifacts. These are invisible bits of formatting left behind when a user copies text directly from a chatbot like Claude and pastes it into their browser. If the tool finds this specific hidden code, it alerts the reader by turning the webpage background bright red.

These digital fingerprints are a byproduct of how modern websites and software interact. When you copy text from a sleek, professional web tool like a chatbot, you are not just copying letters. You are copying the hidden instructions the website uses to tell your browser how to display that text, such as font style or color labels. This is like copying a sentence from a document that has hidden formatting notes attached to it. If you paste that text into another website, those hidden notes can sometimes travel with it. The new detector tool merely looks for one specific label associated with the chatbot Claude. If the label is there, the tool assumes the text came from that AI. However, if a writer uses a different chatbot, or if they copy their work into a standard program like Microsoft Word first — which strips away the hidden formatting — the detector won't find anything at all. This means the tool can't tell if a story is fully written by AI or if an author just used it to check their grammar.

This push for detection puts every writer at risk of being wrongly accused. The community is so protective of human-made art that any sign of a machine is being treated as a betrayal. Because the tools are imprecise, they can easily catch someone who only used a machine to fix a few typos, while missing a full, AI-written story that was simply pasted into a plain text editor first. Ultimately, this creates a culture of suspicion where any writing style that sounds slightly robotic might be flagged as illegitimate. As these tools get more popular, the community faces a difficult question: can you really police creativity without accidentally silencing the very human voices you are trying to protect?