Study finds signs of greater AI use in fiction books on Amazon
A preliminary study and examples of misleading books raise questions about authorship and reliability.
Listen to this article
There is no audio version yet. Request one and an AI voice will read the article aloud.
Read by an AI voice.
.jpg){kind=link}
Researchers have found signs of increasing AI use in self-published fiction books on Amazon. Their study has not yet been peer-reviewed, and the detection software used does not prove who wrote a text.
A study of 14,419 self-published fiction books on Amazon found that the number of new titles has risen sharply since 2022. The researchers linked part of that growth to the spread of language models and found more signs of substantial AI use in the catalogue.
The study was published as a working paper and is therefore not yet a peer-reviewed final publication. The authors used an AI detection tool to classify books. Such tools can provide indications, but on their own do not prove who wrote a text.
The Atlantic also described cases in which AI-generated or allegedly misleading books appeared on Amazon. In one of the cases described, Amazon removed a book from an author profile after the Authors Guild objected.
Amazon allows authors to use AI, provided they follow the platform’s rules. The company does not make every use of AI visible to buyers. As a result, readers often cannot tell whether AI was used only for editing or generated a large part of the text.
The problem affects more than writers. Libraries, booksellers and readers can lose time and money on titles containing fabricated sources, incorrect information or misleading author names. For publishers, an abundance of cheap, automatically produced books could also reduce the visibility of human authors.
The central question, therefore, is not whether AI may ever help with writing, but what information readers receive. Transparent labelling, better checks on author details and clear responsibility for content could make the market easier to understand. Exactly how large the share of misleading books is remains unclear.
Also show the region and a close-up
One story, several perspectives
What is established
- A working paper examined 14,419 self-published fiction books on Amazon.
- The study found a sharp increase in the number of new titles since 2022.
- AI detection can provide indications, but does not independently prove who wrote a text.
Left
Arguments Platforms should protect consumers and human creators from a market in which cheap synthetic content displaces attention and income. Transparency about AI use should therefore be compulsory.
Values Workers’ rights, cultural diversity and consumer protection.
Consequences More rules could make the market more trustworthy, but could also place a heavier burden on legitimate experimental use of AI.
Centre
Arguments AI assistance can be allowed as long as the responsible author and publisher remain identifiable and the content meets the same quality standards as other books.
Values Freedom of choice, responsibility and workable regulation.
Consequences A labelling system could provide clarity without treating all AI use alike.
Right
Arguments Authors and companies should retain room to use new tools; poor books will ultimately be punished by readers, reviewers and the market.
Values Innovation, entrepreneurship and limited platform regulation.
Consequences Overly onerous obligations could disadvantage small publishers, but complete laissez-faire could undermine trust in the market.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved with corrections · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The study and the recent examples can be checked, but the study is not a peer-reviewed final publication and detection software does not prove authorship. That limitation is explicitly incorporated into the text.
- confirmed The study examined 14,419 self-published fiction books on Amazon. — This is stated in the working paper’s summary. source
- confirmed The study is not yet a peer-reviewed final publication. — The publication is listed as an arXiv/working paper. source
- confirmed Amazon removed a book from an author profile after an objection. — The Atlantic describes this example. source
- confirmed AI detection does not prove who wrote a text. — This is a methodological qualification and is presented in the article as a limitation. source
1 correction(s) applied
- Was: AI books are flooding AmazonNow: AI books are taking an increasingly large place on Amazon (‘Flooding’ is too absolute for a study that examined part of the catalogue.)
Editor's note
The study is a working paper and uses AI detection; its results are therefore indications, not a definitive determination of authorship. The recent examples were described journalistically and have not all been independently confirmed by Amazon.Sources
- AI Is Destroying the Trust Between Writers and Readers — The Atlantic
- Generative AI floods and dilutes the market for books — arXiv / NBER working paper
More on this in Dutch media
- Het Parool — „kunstmatige intelligentie”
- NRC — „kunstmatige intelligentie”
- Tweakers — „kunstmatige intelligentie”