AI Book Boom on Amazon
It has been talked about for a long time that artificial intelligence will change the creative industries. First illustration was discussed, then music, video and photography. Now it's the book industry's turn. The rapid proliferation of books produced with the help of artificial intelligence or directly by artificial intelligence on self-publishing platforms such as Amazon Kindle Direct Publishing creates a transformation in the publishing world that is no longer theoretical but can be measured economically.
The study titled “Generative AI floods and dilutes the market for books”, published in July 2026 and updated in August, reveals some of the most remarkable data of this transformation so far. Researchers analyzed the full text of 14,419 self-published genre fiction e-books published on Amazon between January 2023 and March 2026 and matched these books with daily sales data through June 2026. There is an important detail here: The research does not include all books or all e-books on Amazon; It specifically examines self-published genre fiction. So a generalizing statement like “One in five books on Amazon has AI writing” goes further than the research supports. What the study does say is that in the sample studied, books with more than 25 percent of the text classified as AI generated reached a share of approximately 20 percent.
The real striking result is how large the volume of artificial intelligence books entered the market, rather than how well they sold individually. The research suggests that generative AI could transform the publishing industry “with scale, not quality.” In other words, not every book created by AI needs to be a bestseller. The ability to produce hundreds or even thousands of books at very low cost may be enough to change the visibility and income distribution of the total market.
Artificial Intelligence Didn't Make Writing a Book Cheaper; Changed the Economics of Book Production
The traditional production process of a book is quite costly. The writer works for months, sometimes years. The editor steps in. Proofreading is done. The cover is designed. The positioning of the book is determined. Marketing and distribution are planned. Despite all the problems of traditional publishing, this process creates a natural production barrier.
Generative AI eliminates a significant part of this barrier.
Today, a user can produce tens of thousands of words of text in a very short time by giving topic, character structure, chapter plan and tone to a language model. Then, with other AI tools, they can create the cover, write description text, prepare keywords, and publish the book via Kindle Direct Publishing.
This is exactly where the problem begins: While the cost of producing a book approaches almost zero, the reader's attention does not grow at the same pace.
Therefore, the classical economic problem is reversed in the publishing industry. What used to be scarce was content. What's in short supply today is not content, but visibility.
The numbers in the research show this extremely well. While the cumulative catalog examined from the first quarter of 2023 to the first quarter of 2026 grew approximately 38.3 times, the number of books sold within a quarter increased 19.2 times. In contrast, total quarterly revenue grew only 8.9 times. In other words, books entered the market much faster than revenue.
This gap doesn't just mean bad AI books don't sell. The more important finding of the researchers is that the revenue per book also declines in books where AI text is not detected.
Revenue from Non-AI Books Also Drops
Revenue per book decreased in six of the eight genres examined in the research during the comparable sales periods of books published in 2023 and 2025. More importantly, when only examining books with no AI text detected, the decline was seen in seven of eight genres.
This result is important because it alone leaves the explanation that “AI books sell poorly, so it pushes the average down” insufficient.
Researchers define this situation as market dilution.
The logic is simple:
Let's say there are 10 thousand readers and 1,000 books in a category.
After a while, the number of readers increases to 12 thousand, but the number of books increases to 10 thousand.
The market has grown.
But the potential attention per book has decreased.
This may be exactly one of the biggest impacts of Generative AI in terms of publishing. Artificial intelligence does not increase people's capacity to read books; It increases the number of books that can be published.
This means that human writers are now competing for visibility not just with other writers, but with systems that can produce content on an industrial scale.
However, it is necessary to maintain the important methodological caveat of the research. The study itself specifically states that this relationship is observational and does not prove direct causality. So saying “AI books directly reduced the income of so many human authors” would be a stronger claim than the available data suggests. The findings show a strong relationship and market dilution pattern; controlled experimental does not show causality.
The "AI Slop Cannot Be Sold Anyway" Defense Is Not Completely True
One of the common thoughts about books produced by artificial intelligence is this:
“If they are of poor quality, no one will buy them, the market will clear itself.”
The data shows this is only partially true.
Books with more than 25 percent AI text in the research constituted approximately 20 percent of the catalog examined, generating 12.1 percent of sales and 11.3 percent of revenue. Books whose AI text was not detected constituted 62.9 percent of the catalog and generated 72.5 percent of the revenue. So, on average, books that are thought to be human-written have higher performance.
But the story doesn't end there.
According to the study, the share of books that have recently entered the Top 25 lists and have high levels of AI content increased from almost zero to 31 percent during the research period. What's more, one of the highest-earning AI-intensive authors generated roughly $1.7 million in gross revenue before platform outages across eight books; It was reported that a single book produced a gross income of 643 thousand dollars with approximately 80 thousand copies.
Therefore, it is not correct to say “None of the AI books sell”.
The more accurate table is:
Most perform below average, but they gain market share due to very high production volumes, and some examples can achieve serious commercial success.
This is the mechanism that really worries publishers and authors.
While a Human Author Writes One Book a Year, an AI Publisher Can Produce Hundreds of Books
Generative AI's economic advantage is not quality, but marginal production cost.
A human writer cannot work on a hundred novels at the same time.
An AI-based content production operation could theoretically produce dozens or even hundreds of books simultaneously.
When 385 authors who published books with high AI intensity were examined in the research, it was determined that 287 of them increased their monthly publication rate after their first AI book.
This is actually the application of the classic “content farm” model to the book industry.
No single book has to be very good.
100 books are published.
90 won't sell.
Eight less sell.
The two gain visibility in the algorithm.
One of them succeeds.
For the human writer, an unsuccessful book may mean not being rewarded for years of labor. For someone doing AI-based mass production, the same failure is just one of hundreds of attempts at very low cost.
The fact that these two production models compete in the same market naturally creates a serious asymmetry.
Why is Amazon at the Center of This Controversy?
The reason why Amazon stands out here is not because the company produces artificial intelligence. The main reason is that the Kindle Direct Publishing model offers an extremely low publishing barrier.
KDP enables book publishing without the need for a traditional publisher. This system has been an extremely valuable democratization tool for independent writers for many years. An author who was rejected by the publishing house could reach the reader directly.
Generative AI has made it possible to use the same vulnerability for different purposes.
Amazon doesn't pretend to be completely unaware of this problem. KDP's current content policy requires publishers to notify Amazon if they use AI-generated text, images or translations. Amazon also makes a clear distinction between AI-generated and AI-assisted content. If AI directly generated the content itself, Amazon considers it AI-generated, even if it is heavily edited later. The use of artificial intelligence in the author's own text for the purpose of generating ideas, language correction, control or improvement is considered AI-assisted and does not need to be reported separately.
I think this distinction is extremely important for the future of publishing.
Because the debate is “Should using AI be banned?” It's not that simple.
Writing with AI is Not the Same as Having the Book Written by AI
An author to ChatGPT:
“Is there any errors in expression in this paragraph?”
As he asks,
“Write me an 80 thousand word detective novel.”
It's not the same creative process.
Similarly:
“Help me find resources about train stations in 1890 London.”
with
“Create a 30-chapter detective novel in the style of Arthur Conan Doyle.”
There is a serious difference between them.
Amazon's distinction between AI-assisted and AI-generated recognizes exactly this gray area.
Authors Guild also says that it does not completely reject the use of artificial intelligence, but human creativity should be at the center. The organization's approach distinguishes between uses such as grammar checking, research or limited idea development, and the fact that the text of the book is largely created by the model.
Therefore, the healthiest debate of the future will probably arise from the duality of “AI used / did not use AI”.
The real question:
How much of the creative decision belongs to humans?
It will happen.
Amazon Knows How to Use AI, Readers Don't
One of the most critical problems here is transparency.
Amazon KDP requests notification of use of AI-generated content. However, the Authors Guild has long demanded that this information be presented visibly to the consumer. The organization argues that Amazon should make its AI statements available to the reader on the product pages of the books.
This is a very reasonable argument.
A reader might want to know whether the book was created by humans, largely by AI, or in a hybrid form.
After all, we see content information in food products.
We know the fabric ratio in clothes.
Music platforms are starting to tag AI-generated content.
In publishing, "creation provenance", that is, the visibility of the production origin of the content, may become increasingly important.
The aim here does not have to be to ban the artificial intelligence book.
Giving the reader the right to choose may be sufficient.
For example:
Human Authorized
AI Assisted
AI Generated
Even a simple classification like this can significantly reduce the trust problem in the market.
“Human Authored” May One Day Become Valuable Like Organic Product Label
This is exactly why the Authors Guild has created a certification program called Human Authored.
The program aims to verify that the text of the book was written by humans and that the use of AI remains only in limited areas. Certification is not only available to Authors Guild members; Authors of eligible books published in the USA can also participate in the program.
Today this may seem like a small undertaking.
But it wouldn't be surprising if similar tags gained serious marketing value in the future.
Because as AI content increases, paradoxically human labor may turn into a premium feature.
How can:
“handmade”,
“slow fashion”,
“analog”,
“single origin”
If concepts such as these create distinct value in the world of industrial production, in the future the book cover:
Written by a human
The expression can turn into a similar cultural signal.
The better Generative AI gets, the more the value of being human-authored may not automatically disappear.
On the contrary, it may become more visible.
But Identifying AI Books Is Not As Easy As It Is Thought
Here we come to the most controversial part of the research.
In the study, an AI text detection system called Pangram v3.3 was used to determine the AI usage rate of books. By examining the full text of the books, the researchers divided the AI rate into three categories: No AI detected, limited AI up to 25 percent, and high AI use over 25 percent.
But no AI detector is perfect.
Therefore, the study does not say "we have proven with 100 percent certainty that these books were written by AI."
The language of the research is already based on consciously detected AI text.
AI detection systems produce statistical predictions.
It may mistakenly mark human-written text as AI.
Can classify AI-generated text as human writing.
Detection may become more difficult if the text has been heavily edited.
Therefore, blaming an author solely on the AI detector result is extremely problematic.
In 2026, real examples of this began to be seen in discussions in the publishing industry. The Wall Street Journal reports that publishing houses and literary agencies still have not developed a common standard for how to deal with works that suspect the use of AI; reports that some projects have been withdrawn or reviewed due to AI claims.
Here is the big dilemma of the industry:
We want to detect AI text, but the cost of falsely accusing an innocent author of using AI is too high.
Shy Girl Case Shows Publishing's New Problem
One of the concrete examples of this discussion was the novel Shy Girl in 2026. According to the Authors Guild, Hachette conducted an investigation following allegations that the work was written using generative AI and canceled the US publication of the book and ended its UK sales.
The increase in such incidents may lead to a new era of "authorship verification" in the publishing industry.
Publishers will not only evaluate the book.
Maybe he'll want to verify the writing process, too.
Drafts.
Version history.
Editor correspondence.
Research notes.
Document metadata.
All this will lead to the question "Did you really write this text?" in the future. It may become important to answer the question.
This is a very strange time for literature.
For centuries, the problem of the writer was to get his work published.
Now, some authors may also need to prove that they actually wrote the work.
There's a Bigger Copyright Problem: What Are AI Books Trained on?
This is exactly the point where the issue is not limited to new books on Amazon.
Most of Generative AI models learn from human-generated content.
From books.
From articles.
From websites.
From the codes.
The publishing world's objection is therefore two-fold.
First:
Have people's books been used to train AI models?
Secondly:
Are these models now producing new books competing with the same people?
This is exactly one of the strongest objections of the Authors Guild. The organization considers the use of unlicensed books in model training and then the production of content that competes with the works of authors as a serious threat to the creative economy.
New research adds another interesting data to this discussion.
Researchers found that the highest-grossing AI-intensive books included more language patterns that appear in existing books but are rare on the public web. In the 50 highest-earning high-AI books, this rare overlap of expression reaches approximately 45 percent of the text, while in the 50 highest-earning books where AI is not detected, it is 37.7 percent; It was reported that the rate was 19.1 percent for award-winning or award-nominated literary works.
However, we need to be very careful here, too.
This measurement does not prove that a specific work has been copied.
Researchers also state this clearly.
The finding reveals that only high-revenue AI books show higher statistical overlap with rare language patterns specific to existing books.
Still, it is extremely important in terms of copyright debate.
Who Owns the Copyright of a Book Produced by AI in the USA?
Another big legal issue is authorship.
According to the current approach of the US Copyright Office, content created by artificial intelligence does not automatically gain copyright protection just because it is written on a prompt. Sufficient human creative contribution is required for conservation. Human-generated elements, creative editing, or meaningful human modifications may be preserved; However, giving a mere prompt alone is not considered sufficient for authorship.
This creates an interesting paradox for books produced entirely with AI.
One person can create and sell a novel with a few prompts.
But the copyright status of the resulting work may be much more complicated than that of a novel written entirely by humans.
So while AI production reduces the cost of publishing, it makes copyright less legally ambiguous.
Amazon's Problem Isn't Just "Bad Books", It's Discoverability
This is where the real issue becomes a design problem.
Platforms like Amazon are largely discovery systems.
Search results.
Category lists.
Suggestions.
“Customers also bought”.
Bestseller rankings.
Kindle Unlimited.
The algorithm determines which book appears before the reader.
If millions of new content are added to the platform at once, the problem is not storage capacity.
Sorting quality.
For this reason, in the age of AI, Amazon's mission is increasingly moving away from "hosting books".
Main mission:
Understanding which ones are worth showing.
If 15 of the first 20 books in a search result are produced by the same AI content farm, the system may be technically working, but the user experience is unsuccessful.
This is not just a problem for the book industry.
Google is experiencing the same problem with AI-generated websites.
On Spotify AI music.
In YouTube automatic videos.
On Pinterest AI images.
On Etsy AI designs.
The new problem of the platform economy is not the scarcity of content, but the abundance of synthetic content.
How Will Quality Be Measured During the Content Abundance Period?
Before Generative AI, the main problem of digital platforms was producing enough content.
The problem is reversed in 2026.
Producing content is becoming so easy that filtering is increasingly becoming a more valuable technology than production.
We can think of this with this equation:
Formerly:
Value = Ability to produce content
Now:
Value = Ability to produce content + select + verify + position
Therefore, the prediction that publishing houses will disappear completely seems too simplistic to me.
Generative AI could make self-publishing even bigger.
However, as the amount of content on the market increases, the value of reliable publishing house brands as editorial filters may rise again.
A Penguin Random House, HarperCollins or other powerhouse publishing house logo is not just a distribution mark:
“Someone chose this book.”
It means.
During the AI slop, curation may turn into a luxury product.
Book Cover Is Now a Trust Problem
The subject is also extremely interesting in terms of graphic design.
Thanks to generative image models, the cost of creating professional-looking book covers has dropped significantly.
The positive side of this is obvious.
Independent writers can produce better images.
But there is another result:
In the past, low-quality content could often be identified by its low-quality cover.
A bad AI book today that looks extremely professional:
cover,
typography,
mockup,
product description,
even a fake author brand
Can be offered for sale.
So the link between visual quality and content quality is weakening.
This is an interesting ethical problem of design.
Good design can make a bad product more convincing.
In Amazon and other marketplaces, design now has to be part of the trust architecture, not just aesthetics.
Will AI Books End Literature?
Probably not.
The camera did not finish the picture.
Spotify hasn't finished music.
Kindle hasn't completely eliminated the printed book.
ChatGPT will not eliminate the human need to tell stories.
But it may change its economy.
And here is the real danger.
People can continue writing novels.
But if a sufficient number of writers cannot earn income from writing, the field of professional writing may shrink.
U.S. In its evaluations of AI and the creative economy, the Copyright Office also takes into account concerns that the abundance of synthetic content could dilute the income pool of human creators, and points out that the decrease in the livelihoods of creators could affect the amount of human-made works in the long term.
The cultural problem here is not “AI writes bad novels”.
Deeper question:
Will writing good novels remain economically sustainable?
Paradox: The Age of AI May Also Make Human Writers More Valuable
But there is another side to the story.
As the abundance of content increases, the identity value increases.
The reader buys not only the story, but also the person telling the story.
Stephen King's new novel is not just 400 pages of text.
It is part of the product that it was written by Stephen King.
Haruki Murakami.
Margaret Atwood.
Sally Rooney.
Kazuo Ishiguro.
The life experiences, world views and previous works of these names constitute the meaning of the new book.
Generative AI can mimic the same sentence structure.
But it cannot imitate biography.
Therefore, author branding may become more important in the age of AI.
Readers:
Who is this person?
Why did he write this book?
What experience does it come from?
Does it really exist?
He may start asking more questions.
In book marketing, the author's visibility, interviews, process shares, events and personal story can gain greater value.
We Will Experience the Same Thing in Blogs, Design and Advertisement
It would be a mistake to read this issue only through the book industry.
The incident at Amazon is actually an early warning for the entire creative economy.
AI can blog.
AI can make logos.
AI can produce visuals.
AI can produce video.
AI can make music.
AI can write books.
Therefore, production itself becomes cheaper.
For an agency, this means very clearly:
The value of just telling the customer "we produce content" will decrease.
Because content can be produced by anyone.
Actual value:
knowing what content should be produced,
developing original ideas,
Establishing the visual language of the brand,
using the right resources,
verifying the information produced,
conducting quality control,
positioning content within strategy
It will happen.
So AI increases the value of editorial judgment while cheapening the most mechanical parts of creative professions.
“Real Writers Are in a Hard Time” Is True, But Not the Whole Story
The easiest headline for this news:
“AI books invaded Amazon, real writers are running out.”
it would be.
But the real picture is more complicated.
The study indeed shows that AI-intensive books gained a large catalog share, while books without AI detected a decline in revenue in most categories. The pressure this puts on human writers cannot be ignored.
However:
The study did not examine all Amazon books.
AI detection is not 100 percent accurate.
Data alone do not prove causality.
Not all AI books fail.
Not all AI usage can be considered in the same category.
Therefore, the issue is not as simple as “man vs machine”.
The real issue is market design.
How will platforms keep human effort, originality and quality visible in a world where unlimited content can be produced?
The Future of Publishing Maybe Isn't More Technology, It's More Transparency
Amazon's request that AI-generated content be reported to KDP is a step in the right direction. However, if the reader does not see this information, an important part of the system in terms of transparency is missing. The Authors Guild has been demanding exactly this change for years.
Maybe we will see new filters in the book store of the future:
Human Authored.
AI Assisted.
AI Generated.
Verified Author.
Publisher Verified.
Perhaps the "production history" of books will be kept as metadata.
Perhaps digital provenance systems will be used.
None of these are the perfect solution.
But if the reader knows how the content is produced, it can create a healthier market than the use of completely invisible AI.
Conclusion: Artificial Intelligence Made Writing a Book Easier, Not the Value of a Good Book
Generative AI's biggest impact on publishing isn't the fact that it can write books.
Making it economical to produce an unlimited number of books.
This difference is very important.
The number of books people can read is the same.
The hours of the day are the same.
The reader's budget is limited.
But the shelf is now theoretically endless.
Therefore, the problem of the book industry in 2026 is not to produce content.
Finding good content from the crowd.
Amazon's AI book boom tells us something very clear about the future of the creative economy: Originality does not automatically have to become devalued as AI lowers the cost of production.
On the contrary.
When everyone can produce a hundred books, thinking for years and writing just one book can turn into an important story again.
When everyone can create the perfect cover, who wrote the content and why may become more important than the cover.
If anyone can get millions of sentences from a model at once, words are no longer the scarce resource.
The scarce resource is that a person actually has something to say.
And perhaps in the age of artificial intelligence, the most valuable tag of literature will not be “no AI”.
It will be something much simpler:
There is a real person behind this book.
Nur Oğuz