30 August 2026

^{The above image was created by AI. Could you tell?}

 

A future reader may turn over a novel not only to read the blurb, but to check whether its author was human.

That is the idea behind Books By People, a UK organisation placing an Organic Literature stamp on books verified as human-written. The first title announced under the certification was Gonzalo C Garcia’s Telenovela, published by Galley Beggar Press.

The initiative is not simply a protest against artificial intelligence. Its founders want to ensure that human-authored literature remains visible, valued and easily identifiable as generative AI makes the publishing landscape harder to navigate.

“Readers are often the least consulted participants in publishing conversations, despite being the people who ultimately determine which books succeed, which trends emerge, and which stories endure,” co-founder Esme Dennys told LiteraryWorld.

The harder question is what exactly readers should be told.

An author might use AI to check spelling, organise research or suggest alternative titles. A publisher might employ it to produce book metadata. At the other extreme, someone can ask a chatbot to generate an entire novel and make a few cosmetic changes before submitting it.

Both processes involve a person. Few readers would consider them equivalent.

 

AI controversies are forcing publishers to verify authorship

 

Recent disputes over allegedly AI-generated books have turned that ambiguity into a commercial and reputational risk.

In July 2026, agents withdrew Nigerian author Jerry Falade’s debut crime thriller, Call Me, I’ll Hide the Body, after publishers had reportedly offered around $2 million for the rights. Questions had emerged about possible AI involvement, but no conclusive evidence was made public. Falade strongly denied using AI and argued that racial bias contributed to the speed and severity of the backlash.

Earlier in the year, Hachette cancelled the forthcoming US release of Mia Ballard’s horror novel Shy Girl and stopped further distribution after concluding that parts of the book appeared to have been generated using AI. Ballard denied using it herself, suggesting that someone who helped edit an earlier self-published version may have introduced the material.

Then came “The Serpent in the Grove”, Jamir Nazir’s Commonwealth Short Story Prize winner. Readers and AI-detection systems flagged the story as potentially machine-generated. Nazir denied the accusation. After conducting a review, the Commonwealth Foundation said it had concluded that AI had not been used.

Together, the cases expose both sides of the problem. Undisclosed AI can enter the publishing process, but suspicion and software scores can also damage writers who may have done nothing wrong.

Japanese author Rie Qudan offers a more transparent example. She openly said that approximately 5% of her Akutagawa Prize-winning novel Sympathy Tower Tokyo quoted sentences generated by ChatGPT. The AI dialogue was part of a book that itself examined artificial intelligence. Readers could decide what they thought because the author disclosed her method.

 

Can AI detectors identify AI-written books?

 

AI-text detectors generally analyse statistical patterns. Large language models tend to produce relatively predictable word sequences, while human writing is often less regular. Detection software converts those differences into a probability score.

It cannot reconstruct how a manuscript was created.

AI-generated prose can be rewritten until the tell-tale patterns disappear. Human prose can be wrongly flagged, particularly when it is formulaic or written by someone whose first language is not English. A Stanford-led study found that several detectors frequently misclassified essays by non-native English writers.

Books By People consequently decided against making automated AI detection the foundation of its certification system.

“We rejected the idea of using automated AI-detection software, as it is too vulnerable to being deceived and making mistakes,” Dennys told Publishers Weekly.

Instead, the organisation works with publishers and examines their editorial practices, sample titles and evidence of human authorship. Certified books receive an Organic Literature stamp, an identification number and a place in a public directory. Publishers are expected to follow agreed procedures and may undergo spot checks.

The finished prose is only part of the evidence.

 

How can a publisher prove a book is human-written?

 

A genuine book normally leaves debris behind: notebooks, research material, abandoned openings, tracked changes, editorial correspondence and files created over months or years.

Version histories can show a manuscript developing through hundreds of revisions instead of appearing almost complete. Correspondence can document why a character disappeared or an ending changed. Publishers can record which AI tools were permitted and ask authors and editors to declare how they used them.

Together, these records create a human-authorship trail that is harder to fabricate retrospectively. Dennys has described the goal as creating an organic literature market comparable to organic food. A tomato cannot explain how it was grown; confidence depends on standards and records across the supply chain.

The system is not infallible. Publishers earn money from certified titles, declarations can be dishonest and a convincing trail might eventually be fabricated with AI. Its credibility will depend on the strength of its checks, the independence of its decisions and whether certification can be withdrawn when new evidence emerges.

 

Does the EU AI Act require labels on AI-generated books?

 

The EU AI Act approaches transparency from the opposite direction. As MTN explained in Seven myths about what Europe’s new AI rules mean for creators, it does not require every AI-assisted creative work to carry a prominent warning.

Providers of generative AI systems must make synthetic output identifiable in a machine-readable form. Clear public labels are required in particular circumstances, including deepfakes and unreviewed AI-generated text intended to inform people about matters of public interest. Creative and fictional works receive more limited treatment.

The EU rules therefore do not guarantee that readers will find an obvious AI label on every machine-assisted novel. Books By People reverses the proposition: instead of labelling everything artificial, it identifies books produced through a verified human process.

 

Human-written could become its own book category

 

The Organic Literature stamp cannot promise that a novel is original, accurate or enjoyable. Humans have produced terrible books without technological assistance for centuries, after all.

What it can offer is information about how the work was created.

Readers already choose books according to their publisher, translator, edition, paper or environmental credentials. Human authorship could become another useful distinction, particularly in literary fiction and memoir, where the perceived encounter with another person is part of the appeal.

Until recently, human authorship was the default and machine involvement required explanation. Increasingly, human writers may be the ones asked to demonstrate their identity.

The traces of their writing process (drafts, corrections, arguments with editors and abandoned chapters) may become the strongest evidence that they really did write the book.

 

Books connected to human authorship and AI

 

• Telenovela by Gonzalo C Garcia on Amazon or on Bol.com, the first title announced with the Organic Literature certification.
• Sympathy Tower Tokyo by Rie Qudan on Amazon, a novel incorporating disclosed ChatGPT-generated dialogue while examining AI as part of its subject.
• Benny the Blue Whale by Andy Stanton on Amazon or Bol.com, a comic account of deliberate collaboration between an author and ChatGPT.

 

 

Some links in this article may be affiliate links. MTN may receive a small commission if you make a purchase, at no additional cost to you.