It was a statement that had the British book world clutching its pearls. “AI ‘likely’ to produce bestseller by 2030,” read a headline in The Bookseller in June. The story in the parish newsletter of Britain’s book industry reported a speech at a publishing conference by Philip Stone of Nielsen, a company that compiles book sales data in the UK.
The reaction in the normally sedate bookish corners of social media was swift and harsh. Stone’s prediction was, variously, “propaganda”, or “nonsense on stilts”, or “the NFT grift all over again”. The reaction may have been a solid-gold case of shooting the messenger, but it reflected the importance people place on books — novels, stories, non-fiction narratives — as human artefacts.
People thought “I was going to put Lee Child out of a job”, says Stone. “I definitely didn’t say that.” He hadn’t actually said anything about novels at all. “I made what I thought was quite a casual comment [ . . .] for example, could AI produce an adult colouring book? Or a book of ‘dad jokes’ that could sell well enough in that single week before Father’s day to sneak into the top 10.” Only in that sense did Stone “find it likely that AI could produce a bestseller by 2030”.
But the overreaction to his words does show just what a hot and spiky topic artificial intelligence is in publishing right now, for two reasons. The first is that most authors have found that their past works have been ingested by the large language models (LLMs) of AI companies — OpenAI’s ChatGPT is the best-known example — in order to train them to produce text that reads as though it comes from a human brain. The second is that such AI-generated content could supplant human work — and the livelihoods of human writers.
The book world is not the only creative field feeling threatened. Music, art and design are all scrambling to find responses to a technology that can already generate illustrations within seconds or produce cloned voice-a-likes of superstar musicians. Technological developments have always driven fears of human obsolescence. Often these have been overblown. But maybe this time will be different? Big-name artists such as Sir Paul McCartney, Sir Elton John and Dua Lipa certainly think so and have backed high-profile campaigns to restrain AI.
In the literary world, many American authors are suing AI companies, arguing that using their works in this way, without compensation or consent, is a breach of their copyright. In one case in June, the authors lost the argument, with a judge ruling that Meta’s use of millions of books to train its AI systems constituted “fair use”.
“It’s not just frustrating, it’s enraging,” says British novelist Sarah Hall, whose highly acclaimed books, such as Madame Zero and The Electric Michelangelo, have won numerous awards. “It’s not about tools that we’re using — we use technology as writers. It’s about the shift in the balance of trade. Who is going to be getting the money from AI-generated works?” The numbers are not trivial: publishing contributed £11bn to the UK economy in 2024.
Hall is despairing of the British government’s approach to regulating AI companies, and what she sees as its failure to protect the rights of authors. The UK’s Data (Use and Access) Bill — which was finally passed in June after months of parliamentary wrangling — proposes that AI companies should have access to all creative content unless the creator opts out.
“Your books are borrowed from the library and you get a fairly decent royalty from that,” says Hall. “It recognises that this is the vocation and the living of a writer. That Data Bill is toothless. ‘Data’ suggests that this is not art, and it is art, and there’s an ownership aspect to it.”
From left: Sarah Hall, Naomi Alderman and Curtis Sittenfeld © Getty Images; The New Yorker
Naomi Alderman, whose dystopian 2016 novel The Power was adapted for an Amazon TV series, agrees. “I’m not angry about the development of a new and interesting technology. I’m angry about the enclosure aspect, that you’re using everybody’s stuff that belongs to everybody . . . and you get all the money. Now this belongs to these enormous tech corporations, and enormous tech corporations are not good stewards.”
On paper, copyright protection remains strong on both sides of the Atlantic — in the US, creators and their estates are protected for 95 years from the first publication of a work, while in Europe and the UK copyright lasts for 70 years after the death of the author. But for some in the technology industry, copyright, like privacy, is a passing phase. And this is not the only fundamental mismatch between how tech corporations and creators view their work.
Alderman recalls the words of disgraced crypto entrepreneur Sam Bankman-Fried: “If you wrote a book, you fucked up, and it should have been a six-paragraph blog post.” This suggests that tech bosses such as Bankman-Fried, or Mark Zuckerberg, or Elon Musk, don’t really understand that books of the kind written by Hall and Alderman and read by millions are not just vectors for information. They are creative expressions, forms of communication, ways of seeing and thinking.
Which takes us to the second objection that many authors have: that AI-generated books will make them redundant. Why pay an author to write a book when a machine can do it instead? Is this indeed likely, or was Philip Stone right to exclude fiction and narrative non-fiction from his bestseller-by-2030 prediction?
Alderman has tried it. She has long been interested in technology, and writes video games as well as novels and TV series. “It’s almost taboo among writers to say, ‘I am interested in finding out what a person could do with this technology.’” Yet her experiments have not been fruitful. “It hasn’t produced any work that I would have wanted to publish under my own name, because it’s not good enough.”
One of Alderman’s creative outlets is writing scripts for the video game Zombies, Run!, and she tried using AI. “It was so bad!” How? “It doesn’t understand character. It can mimic scene structure, with escalating tension. But we’ve got really deep characterisation, you make a real connection with these audio drama characters. It’s got a heart to it, and AI can’t do the heart.”
US novelist Curtis Sittenfeld, author of American Wife and Rodham, had a similar experience when she published a story generated by AI and intended to be in the style of her own work. As she later told The New York Times, she found the results “boring . . . clichéd, and also shallow in sentiment”.
Anyway, Alderman adds: “Do people want to read novels written by an AI? Is [the purpose of fiction] not to make contact with other humans and to feel less alone?”
This may not be the only reason people read fiction — pure entertainment is a clear purpose — but it is key to the limitations of AI-generated fiction. Hall’s new novel Helm is about the Helm wind in Cumbria, north-west England. “It’s about nature and human interaction and it’s very sensual,” she says. It’s a book that could not be written by AI because it draws on individual human experience. “I have felt the Helm wind. AI has not felt the Helm wind.”
The LLMs, by definition, can only synthesise from existing texts; they cannot originate. Bestsellers and prize-winners tend to come from nowhere, not from everywhere. An AI could never have produced Bonnie Garmus’s blockbuster bestseller Lessons in Chemistry, or Shehan Karunatilaka’s Booker Prize-winning The Seven Moons of Maali Almeida, because nobody was asking for those books to be written. They required human ingenuity.
Yet clearly AI companies would not be putting so much money — and so many of other people’s words — into their LLMs if they didn’t think them capable of writing narrative prose. Sam Altman, chief executive of OpenAI, this year proudly unveiled a short story written by ChatGPT. Notably, it didn’t seek to represent a human experience, being written from an AI machine’s viewpoint. The story was widely dismissed, occasionally praised — for example by novelist Jeanette Winterson — and turned out to have lifted its only interesting phrase (“a democracy of ghosts”) wholesale from Vladimir Nabokov’s novel Pnin.
But within its own limited parameters, the story did its job. Richard Beard, novelist and former creative writing lecturer, tells me: “I’m not 100 per cent certain that if I was given that story by a student, that I would have been able to spot it [as AI].”
Beard believes that if AI does come to replace human-written fiction, “the more formulaic genre novels are going to be the first in the line of fire. A police procedural — you could pretty much feed in [the elements] to ChatGPT. I’ve seen this done: you get a few characters, you feed them in. The thing that I found amusing was it comes back and goes, ‘What quirks would you like your characters to have?’”
Beard, an innovative writer throughout his career, has one plan for how writers can fight back against AI. His new project, the Universal Turing Machine, is an online memoir that he is inviting the public to contribute to. “People remembering their own unique lives” is “the stuff AI can’t replicate,” he says. “With memory, the basic material is not available online.”
As readers, he says, “we like surprises. We like insights we haven’t seen before.” Memoirs can produce this, whereas LLMs are simply statistical models, turning out what they think is a logical and likely form of words. Memoir “remains intensely human, because it doesn’t tend towards the mean”.
For Hall, frustration at the UK government’s approach to AI has led to her own innovation. “The Society of Authors campaigned for AI-generated work to be labelled, but now it’s not going to be.” Earlier this year its US counterpart, the Authors Guild launched a “human authorship mark”. So, Hall says, “Now it’s on us to assert our humanity.” Helm will be published with what she calls a “maker’s mark” on the cover: a small circle containing the words “HUMAN WRITTEN”. “It will assert the craftwork that’s gone into it, the human quality — provenance, authenticity, integrity.”
But if Alderman’s experience with using AI is representative, it may be obvious anyway when a work is AI-generated. “To start out with, I was incredibly impressed with AI-generated images. And then quite quickly, you start to be able to tell what the AI look is. And I think the same has happened with writing,” she says. “It’s got that quality of an auto-correct, where it doesn’t bring you to a conclusion — it just stops. After a while, you stop going. ‘Oh my god, it can do everything’, and you go, ‘Oh, it can do about nine things!’”
AI novels, then, are not with us just yet, but experiments are already under way. This month brings the publication in English of Japanese author Rie Qudan’s novel Sympathy Tower Tokyo, about 5 per cent of which Qudan says was generated by ChatGPT. And, says Nielsen’s Stone: “I think to a certain extent, AI has created books already. When the pope recently passed away, you could go on Amazon within a couple of days and find biographies you could download that were basically scraped from Wikipedia.”
Recommended
Rather than professionally published books being generated by AI, a more likely near-term prospect could be user-generated books that are bespoke to the individual reader. “There might be a system,” suggests Hall, “where you say ‘I want a book about this, this and this’ and you get it.”
“The idea of having words that are written just for you is powerful,” says Alderman. “It can create a very coherent world just for you. But that’s not what novels are for. Novels are for everyone to read, and then we talk about what we see in common or what’s different.”
Are there bright sides to all of this for authors? On the subject of works — including his own — being harvested for AI models, Beard suggests one silver lining. “In some ways, writers should feel very pleased with themselves. We always feel we’re doing something important, despite the [lack of] cultural appreciation and financial appreciation.” He laughs. “And now look! It turns out to be really important, because this is how they train them, this is how they think life works.”
Summer Books 2025
The best titles of the year so far. From politics, economics and history to art, food and, of course, fiction — FT writers choose their favourite reads of the year so far
Join our online book group on Facebook at FT Books Café and follow FT Weekend on Instagram, Bluesky and X