What AI-Generated Books Reveal about Publishing
- David Salariya
- Aug 7
- 6 min read
In August 2025, comedian Rhys James
discovered AI-generated books using his name, likeness, and it was thought to be AI written biography for sale on Amazon, ahead of the release of his memoir. These were not clever parodies. They were said to be sloppy knock-offs, algorithmically generated imposters being sold at suspiciously uniform prices.
The Publishers Association called it a crisis of consumer confidence. The Bookseller ran the story, Amazon issued a lukewarm statement about guidelines and reactive moderation. Everyone looked concerned.
But behind the nervous glances and corporate diplomacy lies a more unsettling truth: AI didn’t cause this problem. It simply made it harder to ignore.
The Real Slop Was Already Here
Amazon’s marketplace has been quietly flooded for years, not just with AI content, but with templated celebrity bios, generic self-help guides, ghostwritten "life manuals," and adult colouring books created with barely a human hand. AI hasn’t lowered standards; it’s simply made their absence undeniable.
We aim to provide the best possible shopping, reading and publishing experience. -
Un-named Amazon spokesperson, The Bookseller, August 2025
What Amazon presents is not a curated catalogue - it’s a
print-on-demand slush pile.
There’s no meaningful editorial oversight, just a jumble of titles dressed up with professional-looking thumbnails and aspirational blurbs. As consumers, we assume books on major platforms come with quality control. Increasingly, they don’t.
Curation Without Accountability
Amazon’s role in the literary ecosystem has quietly morphed from retailer to de facto publisher, without taking on the ethical or editorial responsibilities that role implies. There’s no fact-checking. No attribution standards. Just basic formatting filters and the occasional takedown after complaints surface.
"Our process and guidelines will keep evolving." , Amazon statement on AI books, August 2025
This isn't an oversight. It’s reactive moderation, packaged as quality assurance. Meanwhile, the reader is left to guess: is this real? Is it human? Is it stolen? Who is the author?
But Let’s Be Honest: Trade Publishing Enabled This
While Amazon is rightly in the spotlight, trade publishing cannot absolve itself. In recent years, publishers have quietly:
Removed author and illustrator credits from reprints.
Erased biographies and bylines to foreground branding.
Favoured anonymous house styles over creator visibility.
Publisher and agent in the case of The Salt Path agree who is responsible for fact checking.
Celebrity books being written by ghost-writers - credited and uncredited
These erasures blur the lines between real and synthetic work. If publishers can’t, or won’t, champion authorship, they undermine their own case against AI-generated imposters. We can’t defend the integrity of authorship while also rendering it invisible.
What’s at Stake?
The trust between reader and book has always relied on an invisible contract: someone made this. It was thought through, crafted, curated. That’s the illusion AI-generated content threatens to destroy. But the illusion had already started to crack, thanks to economic pressure, brand-first marketing, and editorial cost-cutting.
AI is not the beginning of the end. It’s just the light that makes the cracks visible.
What Needs to Change?
If the publishing industry is serious about protecting writers, readers, and the credibility of books, it needs more than polite concern.
1. Mandatory Attribution
Every edition of a book, print or digital, must list its creators, visibly and accessibly.
2. Platform Responsibility
Retailers like Amazon must adopt genuine editorial oversight. Automated filters are not enough.
3. Transparency in Publishing
Publishers must stop erasing credits, hiding packagers, and prioritising anonymity over authenticity.
4. Cultural Literacy
We must teach readers how to spot the signs of AI-generated content, so they can choose with confidence.
The current wave of AI-generated “slop books” on Amazon has clear roots in a pre-AI problem identified as early as 2022, when the Publishers Association warned of so-called “leech publishers” uploading unauthorised workbooks and summaries based on newly released titles. These knock-off publications, often badly written with a few pages, mimicked real books with titles like Workbook for You Are Not Alone or Shortnotes on Down and Out in Paradise, yet cost nearly as much as the originals. Despite vague disclaimers, they were designed to mislead consumers. At the time, Amazon removed listings only after complaints, and issued refunds upon request, policies still in place today. As Claire Anker of the Publishers Association noted then, these self-published imitations were exploiting author labour and often infringing copyright. In hindsight, this was not a glitch in the system-it was a warning of what was to come.
I wouldn't say this is just a skirmish over e-books. It’s a test of the publishings values. Do we believe authorship matters? Do we think readers deserve to know who made what? Or are we content to drown in a slurry of synthetic stories and call it progress?
If the book world is to hold the line against the worst of AI content, it must first look in the mirror. We don't just need better tech - we need better ethics.
When Licensing Looks Like a Land Grab
The proliferation of “AI slop” on Amazon is one problem, Bloomsbury has now confirmed that it is “exploring” licensing its backlist to tech firms for AI training purposes, offering authors a modest 20% slice of any bulk deal that materialises.
Yes, 20%.
That’s one-fifth of the revenue, not from individual negotiations, but from a blanket agreement covering multiple titles, with no transparency on pricing or on which AI models are being trained, for what purpose, and to what end. Authors who co-wrote books are to receive a “pro-rata share,” potentially dwindling further. And authors who don’t opt in? Well, they’re now faced with the unspoken suspicion that they’re the ones “blocking progress.”
This isn’t partnership. It’s pre-emption.
Who Owns AI Training Rights?
As the Society of Authors (SoA) rightly points out, AI training rights are new rights. They are not automatically granted to publishers unless explicitly negotiated. Licensing them retroactively through opt-in schemes is not an act of benevolence; it’s an attempted legal workaround disguised in the language of innovation.
What is the publisher doing for their 80% share?” - Anna Ganley, CEO, Society of Authors
It’s a fair question. Unless Bloomsbury is personally programming the large language model or offering indemnity insurance for reputational fallout, the lion’s share seems speculative at best, opportunistic at worst.
Compare this to HarperCollins, which offered a 50:50 split and a fixed $2,500 fee per book for a three-year licence. That at least resembled a transaction. Bloomsbury’s model resembles a content sweep: wide, vague, and strategically opaque.
The Illusion of Consent
What’s more troubling is the rhetorical sleight of hand. Bloomsbury’s spokesperson described licensing as “an important means of protecting and enforcing copyright.” On paper, yes. But when the only protection offered is participation in a scheme whose details remain undisclosed, it ceases to be protection and starts to look like capitulation.
A viable licensing ecosystem... supports the case that the current copyright regime does not need to change. - Bloomsbury spokesperson
But that statement collapses under scrutiny. If publishers claim to be protecting copyright while quietly repackaging rights authors never granted, they are reinforcing the very erosion they claim to resist.
The Bigger Picture: Ethics Disguised as Strategy
Bloomsbury also claims this initiative is part of the “ethical development of the world’s future knowledge ecology.” It's a grand phrase, and perhaps sincere in intent. But let’s not mistake the ethics of participation with the ethics of ownership.
The core issue isn’t whether authors should ever license their work for AI training. Some might choose to.
But they should know:
What model is being trained
What their work is contributing to
Whether their style or ideas will be cloned
Whether derivative content will be monitored, shared, or sold
And who, if anyone, will be credited, compensated, or consulted
Until publishers can provide those answers, they should not be offering the illusion of informed consent.
The Future Is Licensed… But By Whom?
If tech companies want to train models on high-quality material, they should pay fair value and negotiate with the originators of that content - not siphon rights through bulk deals struck behind closed doors.
Likewise, if publishers want to position themselves as ethical stewards of culture, they must stop thinking like rights aggregators and start behaving like creative partners.
Until then, the publishing world is not licensing knowledge. It’s licensing away trust.
About the Author
David Salariya is an author, illustrator, book creator, and designer who has spent a lifetime creating books that children wanted to read. As the founder of The Salariya Book Company, he produced award-winning children’s non-fiction series including You Wouldn’t Want To Be… and A Very Peculiar History. His work has been published in over 45 countries and translated into more than 35 languages.
He has stood on the edge of Victoria Falls, been chased by an elephant, and once published a book that sold 40 copies. He has also published titles that sold in their millions.
David is now writing, painting, and developing a new book series Pulp History, a series that digs into the overlooked, outrageous, and occasionally grotesque episodes of the past - served with sharp storytelling and a splash of irreverence.
He lives in Brighton, England, and continues to campaign (quietly but firmly) for the radical idea that authors, illustrators and book creators should be
credited for their work. His work from the last 40 years is being republished by Hatch an imprint of Bonnier Books. www.davidsalariya.com