AI Books Aren't Just Adding More Content. They May Be Changing the Economics of Publishing
For the last couple of years, much of the discussion around AI-generated books has centred on quality.
Can AI write a book that readers genuinely enjoy? Will readers be able to tell the difference? Will human authors always have the advantage because lived experience, judgement and originality cannot simply be generated?
Those are interesting questions, although I don't think they are the most important ones anymore.
New research suggests the real impact of AI on publishing may have less to do with whether AI books are better and much more to do with what happens when the cost of producing another book falls close to zero.
Researchers analysed 14,419 self-published genre-fiction ebooks sold on Amazon between 2023 and 2026, comparing AI-generated text with actual sales performance. They found that books containing substantial AI-generated text now make up a significant proportion of the catalogue and are increasingly appearing in commercially successful positions, even though they generate less revenue per title than books with no detected AI text.
The statistic that stayed with me was this.
During the period studied, the number of books recording sales in a quarter increased 19.2 times, while total quarterly revenue increased only 8.9 times.
In other words, the market added books much faster than it added readers' spending.
That means the average revenue available to each selling book became smaller. The researchers describe this as a dilution effect, where the growing volume of books changes the economics of the market even if individual AI-generated titles are not dominating sales. Their conclusion is that generative AI can reshape a creative market through scale rather than quality. (arXiv)
I think that distinction is incredibly important.
For a long time, many of us have assumed that poor-quality AI books would simply disappear because readers would ignore them.
This research suggests that may not be how markets behave.
If publishing another title becomes almost free, then simply increasing the number of books changes the competitive landscape. Even if many of those books are fairly ordinary, they still compete for visibility, recommendation algorithms, bestseller rankings and reader attention.
Business Insider highlighted this research because it provides evidence for something authors have been arguing for some time. The debate is no longer only about whether AI companies trained their models on copyrighted books. It is increasingly about whether AI-generated works can have a measurable commercial impact on the authors whose work helped train those systems in the first place. The research does not settle that legal question, although it does provide fresh evidence that market dilution is becoming a real phenomenon. (Business Insider)
I do think it is important to keep the findings in context.
This is currently a working paper, not a peer-reviewed journal publication. The researchers also focused on self-published genre fiction sold through Amazon rather than the entire publishing industry. Like all AI detection research, it relies on the accuracy of current detection methods, which continue to evolve.
Even with those qualifications, I think the broader direction deserves attention.
The question for expert authors may no longer be:
"Can AI write a book like mine?"
A more useful question could be:
"If thousands of additional books can be produced at almost no cost, what makes my work valuable beyond the manuscript itself?"
That is where I think the conversation becomes much more interesting.
I'm not especially excited by the idea of using AI to manufacture larger quantities of books.
There is already more information in the world than any of us could read in a lifetime.
What interests me is whether AI can help readers gain more value from work that already exists.
That is one of the reasons I have become so interested in AI Book Companion™.
Instead of asking AI to produce another book, I would rather use it to help a reader explore the thinking inside an existing one. They can question a framework, understand how it applies to their own situation and return weeks later when they are trying to implement what they have learned.
The AI isn't replacing the author's work.
It is extending the relationship between the reader and the author's intellectual property.
Those feel like very different uses of the same technology.
If the economics of publishing continue moving towards abundance, I suspect the long-term advantage will belong less to the people who can produce the greatest volume of content and more to the people who create distinctive thinking that readers want to spend time with.
A book may still be the starting point.
It simply doesn't have to be the end of the experience.
Three Key Insights
- New research suggests AI-generated books can reshape publishing through volume, even when individual AI books do not outperform human-authored books.
- The number of books competing for readers is growing much faster than reader spending, making discoverability an increasingly important challenge.
- For expert authors, the greatest long-term value may lie in the thinking, frameworks and reader experience that sit around a book rather than the book alone.
Three Questions to Consider
- If producing another book becomes almost free, what makes your work genuinely difficult to replace?
- Is your book currently the finished product, or could it become the starting point for a deeper reader experience?
- As AI increases the supply of content, will readers place a higher value on trusted expertise and original thinking?
Sources
Business Insider
Evidence of unfair use: AI books squeeze human authors out of the market
Published: 15 August 2026
https://www.businessinsider.com/ai-books-anthropic-training-human-authors-market-amazon-research-2026-8
arXiv
Generative AI floods and dilutes the market for books
Tuhin Chakrabarty, Xinyue Liu, Jane C. Ginsburg & Paramveer Dhillon
Published: 22 July 2026 (Working Paper)
https://arxiv.org/abs/2607.20349