Three years ago, in June 2023, I was scrolling through the Amazon Kindle bestseller list. All the regular, normal, to-be-expected titles were at the top. But as I kept scrolling, the books began to change.
Gone were the big names and publishers. Gone were the slick covers and compelling titles. Instead, I found myself faced with titles like Department of Vinh Du Stands in Front of His Parents’ Tombstone and Apricot Bar Code Architecture. ChatGPT had debuted six months prior and folks were just figuring out that LLMs could be used to create AI-generated novels.
Not that they were good at it. Apricot Bar Code Architecture began: “Black lace pajamas, very short skirt, the most important thing, now this lace pajamas are all wet.”
Thrilling reading, to be sure. And — until Amazon caught wind of it and took it down — it could be yours for only $8.90 or you could read for free with a Kindle Unlimited subscription.
“Here’s how the scam works.”
If you’re wondering who’s buying such obvious AI slop, it’s the same folks who created it in the first place. Here’s how the scam works.
- Get the AI of your choice to write a novel.
- Use AI to create a cover or use the stock covers available on Amazon’s Kindle Direct Publishing software.
- Use Amazon KDP to upload your book on Amazon. At present, Amazon asks you to tick a box if any part of your book is AI-generated or assisted, but it’s all on the honor system.
- Make the books available on Kindle Unlimited, a subscription service that allows readers library-like access to books for a monthly fee. Authors get paid a variable royalty based on a number of factors, but on average are paid about half-a-cent ($0.004537) per page.
- Engage click farms and AI bots to “read” the books on KU. A bot that is able to read 100,000 pages per month will generate between $400 and $500 per month income. In Vietnam, where many of these AI-generated books originated, that was greater than the median monthly income.
Once it became known, the scam didn’t last long. Amazon took down the books and put guardrails in place to detect bot-run KU accounts. But AI-generated literature persists and in the years since AI has only gotten better.
(Shutterstock)
The AI-generated book boom
Since ChatGPT was released as the first widely-available LLM in November 2022, the number of e-books published each week has tripled, with detected AI use also rising significantly during that time. AI generation drastically lowers the cost, time and ability needed to produce a book. Combined with the accessibility of digital self-publishing, producers of AI-generated material — to call them “authors” would be to give them too much credit — can quickly and easily generate new books at a pace that puts even James Patterson to shame.
Take, for instance, the author Coral Hart. A real person and established author within traditional publishing, Hart (a pseudonym, not her real name) would release 10 to 12 books per year. Now, with AI, she’s producing more than 200 — earning what she claims is “high six figures in revenue.”
That claim shows the financial benefit of quantity. In an interview with the New York Times, Hart revealed her books had sold 50,000 copies. At 200 novels, that’s an average of only 250 copies sold per title. Hardly bestseller material. But the cumulative effect is nearly a million dollars of sales in one year.
During the interview, Hart showcased her novel-generating capabilities, composing a whole book over a 45-minute period. In addition to publishing AI novels, Hart supplements income by training others to employ her same methods.
The cost of infinite books
This influx of AI-generated literature has major repercussions for authors creating actual stories. An August 2026 research study found the explosion of AI-generated books into the general market diluted the market, resulting in decreased revenues for all books. The rapid onslaught of AI-generated books effectively crowds out human-generated books.
AI-generated titles compete within the same Amazon algorithm as books written by actual authors. An AI producer can publish a hundred books a month, focused on key niches and in-the-moment marketing, buy fake reviews and create enough traction to push legitimate titles down the search result rankings. As Walter Schulze writes, “The authors who built their livelihood on (Kindle Direct Publishing) are now competing against what is essentially an inventory machine rather than a publishing market.”
Compounding the indignity is that LLMs gained the ability to generate full-length novels through being trained on unlicensed copies of copyrighted works. In September 2025, Anthropic — which owns Claude — reached a $1.5 billion settlement with a group of authors and publishers after a judge ruled they had illegally downloaded and stored millions of books on which to train their AI model. Ask a generative AI model to write a paragraph in the style of Dr. Seuss and it’ll know exactly what you mean because it has every Dr. Seuss book within its knowledge base.
“Those producing AI-generated works are relying on the model’s training from actual authors.”
Those producing AI-generated works are relying on the model’s training from actual authors. For example, Darkhallow Academy: Year 2 by Lena McDonald was outed as AI-generated when the author accidentally left in an LLM response stating, “I’ve rewritten the passage to align more with J. Bree’s style.” J. Bree is an actual human author of an internationally bestselling series of romance/fantasy.
The rise in AI-generated novels has caused 10-time USA Today best-seller Jess Michaels to consider giving up writing. Talking to The New York Times, Michaels said, “The glut of the AI work makes discoverability very difficult. … I cannot tell you how soul-crushing it is. It essentially erases actual authors.”
This problem isn’t just relegated to romance and/or fantasy of the rather risque variety, either. It affects non-fiction as well. Multiple non-fiction authors have found their work summarized or duplicated by AI-generated books. Author Marie Arana told NPR she invested four or five years of research into her book Latinoland and the day after it published Amazon had an AI-generated Summary of Latinoland for sale.
Big name authors aren’t the only ones at risk. Two years ago, occupational therapist Kelsie Olds (my wife) self-published a book based on her work in child advocacy called Your Child’s Point of View: Understanding the Reasons Kids Do Unreasonable Things. It was briefly No. 1 in its Amazon niche of family/parenting. A few weeks later, an AI-generated knockoff with a similar title was below Kelsie’s in an Amazon search. It has since been removed.
When it comes to non-fiction, AI-generated titles are also prone to misinformation. Tech journalist Kashmir Hill found herself the subject of an AI-generated biography of herself. In her words: “My biography is 90 pages long and should be shorter. … (It) is flattering, fabricated and absolutely packed with em dashes. It bears all the signs of AI slop.”
The speed of AI generation allows AI producers to capitalize on current events. Within hours of Charlie Kirk’s assassination, Amazon was flooded with AI-generated titles capitalizing on his death. One title, The Charlie Kirk Shooting: A Nation on Edge, claimed to give readers the full story, including the “quick arrest of the Charlie Kirk shooter.”
At the time it was published, Kirk’s alleged shooter had not yet been arrested. Another, The Shooting of Charlie Kirk, appeared to be published the day before Kirk’s assassination, fueling conspiracy theories about the attack.
(Shutterstock)
The end of obvious AI
As AI-generation improves, so does its climb up the bestseller lists. The days of obvious AI slop like Apricot Bar Code Architecture are gone. We now have tens of thousands of full-length books that can gather an actual fanbase and be positively received, particularly if they are within highly specific niches.
In February, Simon & Schuster paid a seven-figure sum to secure the rights to the viral self-published novel Daggermouth and its not-yet-written sequel. Daggermouth hit the USA Today best-seller list and boasts nearly 107,000 ratings at an average of 4.33 out of 5. Some AI researchers are convinced that it’s at least partially AI-generated. Daggermouth’s author, H.M. Wolfe, denies the claims.
“It’s gotten to a point that “good writing” is seen as a sign of AI.”
And we should be skeptical of AI detectors. Instances abound of detectors falsely flagging AI, likely because the AI is trained on actual writing. It’s gotten to a point that “good writing” is seen as a sign of AI. The truth is that we just aren’t very good at determining the difference between human-generated and AI-generated text. We are hardwired to believe communication in human language came from other humans for the purpose of connection.
AI-generated literature is no longer a laughable attempt to scam Amazon, but a serious attempt to scam, or at least deceive, readers. Life is story. Authors — fiction and non-fiction — share themselves with us when they write. We get an intimate look into their mind as we consume their work and get drawn into their art. AI-generation breaks this human connection, for the purpose of making money with the least amount of effort.
It’s something George Orwell saw the danger of decades ago. In his novel 1984, the Ministry of Truth has “a whole chain of separate departments dealing with proletarian literature, music, drama and entertainment generally.” The protagonist’s love interest, Julia, works “on the novel-writing machines in the Fiction Department” and that, in this dystopia, “Books were just a commodity that had to be produced, like jam or bootlaces.”
It’s a shockingly accurate description of modern AI-generated literature. It steals our humanity and gives it back to us in a sanitized, optimized, algorithmized format. It colonizes artistic expression and human connection, reducing its value to a mere economic transaction.
Three years ago, I scrolled past AI-generated books on Amazon and had a good laugh at how bad they were. Today, I’m no longer laughing.
Josh Olds
Josh Olds is a public theologian and pastor for those disillusioned with institutional church. He is the creator of the small-group video series “Year on the Mountaintop” and a featured contributor to Fostering Hope: A Prayerbook for Fostering and Adoptive Parents. Follow his work on Facebook or at JoshOlds.com.
Related:
Do you hear what I hear? | Analysis by Rick Pidcock
What AI worship exposes about our songs and our souls | Analysis by Josh Olds