Authors Can Defend Against AI Allegations With ‘Paper Trail’ Tools. Is It Enough?

The publishing industry's AI paranoia has acquired a new target: the authors themselves. In March, Hachette Book Group canceled a US publication run for Shy Girl, a horror novel by Mia Ballard, and discontinued the UK edition, where the book had already been released. The decision followed widespread social media speculation and a Times article questioning whether the novel was machine-written; Pangram, an AI detection service, judged some 78 percent of the text AI-generated.

Ballard has denied using AI to write the book, saying an editor was responsible for AI-assisted changes to a previously self-published version, and she says she is pursuing legal action. Several other fiction writers have faced similar suspicion in quick succession — one after winning a short story prize, another after publishers bid into the millions for his crime novel — and both have denied the claims. Andrea Bartz, lead plaintiff in the class-action suit against Anthropic, says the accusations land long before any rebuttal can: "I kind of immediately felt like this was going to lead to a potential witch hunt." The unavoidable arrival of AI in publishing, she argues, means anyone can accuse anyone, and "There's no agreed upon standard" for proving human authorship.

Guild CEO Mary Rasenberger nonetheless calls the exercise "insulting" and a "burden," saying authors should not have to preserve receipts. She contends the economics cut against AI use — "What you're selling are your words and your voice and your imagination" — and reports that editors of literary fiction and nonfiction say they rarely encounter AI-generated submissions, some claiming they would recognize it immediately. She also cautions against accusing fellow authors, which she says does not support a healthy literary community, and warns that detection tools can be fooled and produce false positives.

That caution cuts to the core of the standoff. Pangram claims an industry-leading false positive rate of 0.01 percent on creative writing, yet Rasenberger says the Authors Guild has "heard of lots of instances" where authors fed their own human-written work into Pangram and received an AI verdict, with various factors able to trigger misclassification. What remains open is whether paper-trail tools and detectors can converge on any shared evidentiary standard — and whether publishers, prize committees, or courts will treat such records as proof, rather than shifting the burden of suspicion back onto writers.

keyFactsAndrea Bartz has published five thrillers since 2019, with options from Netflix and Hulu.

Bartz was lead plaintiff in the Anthropic class action; it settled for $1.5 billion.

Pangram is characterized as robust but by no means infallible.