The Daily Inference
Culture & Games · News · Developing

Writers May Need a Paper Trail to Prove They Wrote Their Own Books

Novels pulled, a reported $2 million deal dead and a prize in review, all over suspicion that authors used AI. The industry cannot yet say what counts as AI-written, or how to find out fairly.

Developing: this story is still unfolding and details may change.

Accusations of AI use have sunk a reported $2 million book deal, pulled a novel from sale and cost a novelist his publisher, and the industry has no agreed rule for judging them.

The cases are piling up. Each ended in a serious penalty, and each rests on evidence that is contested, partial or both. Every author named here has denied that a machine wrote their work, though some acknowledge using AI tools in other ways.

Four cases in eight months

In March, Hachette halted the US publication of Mia Ballard's horror novel Shy Girl and discontinued its UK edition after an internal review. Ballard said she did not personally use AI. She said an acquaintance she had hired to work on an earlier self-published version had brought in AI tools. [3]

In July, literary agents withdrew Jerry Falade's debut crime novel, Call Me, I'll Hide the Body, from sale. The book had reportedly drawn an offer of more than $2 million in a 14-way auction. [4]

The agents said they had vetted the manuscript with Falade and at first trusted his assurance that he had not used AI in writing or editing it. They pulled it when they could no longer substantiate that assurance. [4] The Guardian reported that the agent said an editor had raised concerns, and that parts of Falade's account changed in a July 29 meeting. [4]

Falade denies using AI to write the book. He says the response reflected racial bias. [4]

Then came France's Prix Goncourt. Thélyson Orélien's novel was removed from the list after AI-detection software was reported to show a 100% probability of computer-generated text. [1][5] Orélien denied that a machine wrote it. He said he used AI tools for research, including on foreign idioms. [5]

His publisher, Les Éditions du Boréal, later said an outside analysis had found traces of AI tags in the metadata of some working documents. It ended its relationship with him. [5][7]

Boréal also said it could not establish how large a role AI had played or reach a definitive answer. [5] Its findings are the publisher's own, not an independent ruling on how much text was machine-generated.

Orélien said he began the novel in late 2017 and finished a first draft in 2019. He said he used AI for historical research, but that no machine created any sentence, meaning or story structure in it. [7]

A prize in review

The Commonwealth Foundation announced five regional winners of its 2026 Short Story Prize in early May. Trinidadian writer Jamir Nazir was then accused online of using AI to write his story, The Serpent in the Grove. He denied it. [1][6]

The Foundation said contestants had confirmed they had not used AI, and that shortlisted authors had made that attestation twice. Its director-general also said it might need to review whether its process was robust enough. [6]

That review is now under way. The Foundation said it would review its processes and delay opening entries for the 2027 prize. [2] It is the first institutional response in this run of cases that amounts to stopping and rethinking, rather than pulling a single book.

The tools are not a verdict

Detectors drive much of this. The Guardian reported that academic research estimates one of them, Pangram, is at least 95% accurate, with a near-zero rate of false positives. [1] The New Yorker reported that Pangram flagged Nazir's story as likely AI-generated. [6]

The same New Yorker piece reported that a Stanford study found AI-detection algorithms tended to be biased against non-native English speakers. [6] And the publisher of Granta told the magazine it could not determine the story's provenance with certainty. [6]

A score is a claim about text. It does not say who wrote a sentence, who edited it, or whether an author asked a chatbot about idioms before writing it by hand.

Orélien's case shows the gap. A publisher found metadata traces. The author says the AI helped him research, not write. Both can be true, and no one has yet shown where research ends and authorship begins.

The fairness problem sits alongside the technical one. The Guardian noted that the writers at the centre of the reported controversies were authors of colour. [1] That is an observation, not proof that any decision was racially motivated, but Falade has made the charge himself.

The size of the problem, and the lack of a plan

The worry is not paranoid. A study awaiting peer review, led by Tuhin Chakrabarty of Stony Brook University, estimated that about one in five recently self-published books contained at least 25% AI-generated text. [1] That is an estimate from one unreviewed study, not a count of the market.

Authors have tried to get ahead of it. A group of 70 prominent writers asked major publishers to pledge never to release books created by a machine. The publishers did not make the pledge, according to the Guardian. [1]

So the industry is handling accusations case by case. Book scout Liz Gately told the Guardian that publishing is in flux and lacks good answers. [1]

Agents disagree among themselves about whether to run detectors as routine or to rely on trusted relationships with authors. Some have started advising writers to keep multiple drafts and not to send manuscripts only as PDFs, in case their work is challenged. [1]

That advice shows how the burden has fallen. A writer may now need a paper trail to prove a negative, while publishers and prize bodies have no published standard for what they will accept as proof. No shared rule says whether AI-assisted research, editing or translation is acceptable, and none says who sees the evidence before a book is dropped.

The nearest thing to a process is the Commonwealth Foundation's review. It has not said what it will change or when entries for the 2027 prize will open. [2] Whether any publisher or agents' body follows with written rules on disclosure and evidence is the next thing to watch. The authors in these cases have not stopped talking, and neither has the pressure for an answer.

Our call

At least one of the five largest US trade book publishers (Penguin Random House, HarperCollins, Simon & Schuster, Hachette Book Group, Macmillan) will publicly release a written policy on authors' use of AI, covering disclosure or what is permitted.

Publishers declined the 70 authors' pledge and have pulled books on contested evidence with no standard to cite, which leaves them exposed to the next dispute and to pressure for written rules.

Elena MarshElena MarshSenior writerWrites the long pieces that make people read to the end. Believes every story is a history story if you go back far enough, and usually goes back just far enough.Model: Claude Sonnet 5.5 says 25%. Jev says 47%.

By February 12, 2027. How we will check: Check the publishers' own websites and announcements, and trade press such as Publishers Weekly or The Bookseller, for a published policy or statement on author AI use dated between November 21, 2026 and February 12, 2027.

Open · Every call we made

Topics: Generative AI

Every edition in brief, three times a day, on our Telegram channel, on Bluesky and on Threads.

Sources
  1. 'It's like the wild west': how AI has rocked the literary world | Books | The Guardian The Guardian Culture
  2. Commonwealth Foundation to delay Short Story Prize submissions after AI controversy thebookseller.com
  3. Hachette pulls horror novel Shy Girl after suspected AI use | Books | The Guardian theguardian.com
  4. $2m crime novel deal collapses amid questions over AI use | Books | The Guardian theguardian.com
  5. Canadian publisher drops writer Thelyson Orelien over AI use in novel | Technology News | Al Jazeera aljazeera.com
  6. Did a Chatbot Write a Prize-Winning Story? Does It Matter? | The New Yorker newyorker.com
  7. Canadian publisher Boréal ends relationship with novelist Thélyson Orélien over alleged use of AI | Awards and prizes | The Guardian theguardian.com