Where does publishing’s AI problem leave authors and readers? - Daily Press

The publishing industry is grappling with a surge of distrust as artificial intelligence increasingly influences book creation, leaving authors and readers uncertain about the authenticity of literary works. This climate of paranoia reached a boiling point in March 2024 when Hachette, a major publisher, cancelled the release of the horror novel Shy Girl by Mia Ballard in the United States. The decision followed evidence suggesting the book had been produced, at least in part, by AI. Hachette subsequently pulled the title in the United Kingdom, where it had previously been released following Ballard's initial self-publication.

What happened

The controversy surrounding Shy Girl highlights the significant challenges publishers face in vetting manuscripts for AI-generated content. Before the publisher intervened, readers on platforms like Goodreads and Reddit had spent months identifying and complaining about what they characterized as obvious evidence of chatbot language within the book. The incident has raised critical questions regarding how major publishing houses evaluate the work they acquire and whether they are equipped to detect synthetic prose.

In the wake of the cancellation, many authors have expressed concern that the industry is shifting toward an era of intense scrutiny. Writers, including debut novelist Antonio Bricio and thriller author Andrea Bartz, have reported that even human-authored work is being incorrectly flagged as AI-generated by detection software. These tools have demonstrated inconsistent results, with minor edits sometimes causing a program to flip from 100% AI-generated to 100% human-written.

While major publishing houses currently lack standardized rules regarding AI, the industry remains divided on how to regulate the technology. Some authors are turning to third-party certifications, such as a “human authored” logo offered by the Authors Guild, in an attempt to provide transparency to readers. However, the lack of consensus and the unreliability of detection software have left many writers feeling vulnerable to unfounded accusations, while readers increasingly demand the ability to distinguish between human-penned stories and machine-generated text.