The books were never the business
A novelist’s 370 AI-assisted titles were a demonstration, and the method that produced them is what she now sells to other writers.
Sign up for my Substack daily AI newsletter https://johanosteyn.substack.com
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
A South African romance novelist writing as Coral Hart has published more than 370 books since February 2025 with the help of artificial intelligence, under more than 10 further pen names. By her own estimate she has sold over one million copies and earned over one million dollars this year, and one novel reached Amazon's top 100 bestsellers. The argument that followed was about whether any of it counts as literature. That argument misses the more interesting fact in the story. She is preparing to launch AI-writing software of her own. The books were the demonstration, and the method that produced them is now the product. For anyone who makes things for a living, that sequence is worth studying closely, because it describes what happens to a creative market once the cost of producing the work falls to almost nothing.
CONTEXT AND BACKGROUND
The flood is measurable. Monthly new ebook releases averaged roughly 100,000 between 2020 and 2022, and rose above 300,000 a month by late 2025. Amazon has responded by rationing shelf space rather than judging content. On 21 September 2026 Kindle Direct Publishing cut authors from ten new titles a week to two per book format each week, saying it wanted to maintain a positive publishing experience. The company did not say the change was about AI, although the trade reads it that way.
Hart is not an outsider to the industry she is accused of degrading. She worked in publishing for 14 years, sets the concept, characters and chapter outlines herself, uses AI for developmental editing and continuity, and sends the result to a human proofreader.
INSIGHT AND ANALYSIS
Follow the money and the story changes shape. An advantage built on producing more of something cheaply cannot last, because the same tools reach every competitor within months. What survives is the workflow, which is exactly what she has begun to sell.
Economic research points the same way. Modelling by the Federal Reserve Bank of San Francisco finds the effect of AI on concentration is not a straight line. While AI use is low, large incumbent users gain share. Once adoption broadens, smaller adopters take it back, and the research concludes concentration is likely to fall as adoption rises. The window in which volume pays is closing by design, not by regulation.
I have previously written about this in Everyone Bought The Tools, Almost Nobody Changed The Work, where the finding was that organisations gaining from AI did so by rebuilding a process rather than buying better software. Hart rebuilt the process of making a novel. That, rather than the novels, is the asset.
The counter-arguments are substantial. Every figure describing her scale comes from her, unaudited, and a sceptical reader may treat the account as marketing for the software she is about to sell. Tooling markets also commoditise faster than the output they serve, so the method's seller can be undercut as fast as the book's seller. A bifurcated market is also possible, in which readers who dislike machine-made work pay a premium for writing a person produced. Read generously, the central bank finding suggests the flood is new entrants competing, which makes the complaint about low-quality work the familiar sound of incumbents objecting.
IMPLICATIONS
The limit that bites first is not being set by critics, ethics bodies or copyright law. Those may reshape the field more profoundly in time, and none moves at the speed of a platform policy change. The limit in force is distribution. Deezer's chief executive says half of all daily uploads to the service are AI-generated tracks, peaking in June 2026 at a monthly average of 90,000 tracks a day. Etsy has prohibited AI-generated patterns, recipes and instruction manuals along with bot-generated listings, and now requires one unedited photograph of the finished item, saying many buyers can no longer tell which products involve AI.
A better test than quality already exists. The Academic and Non-Fiction Authors' Association of South Africa, framing its April symposium around publishing moving from AI experimentation to AI integration, cited the Committee on Publication Ethics position that AI cannot be an author because it cannot take responsibility. Accountability survives the flood in a way that arguments about taste do not.
For anyone running a creative business, three questions follow. What do we own that cannot be packaged and resold. How much of our access to customers rests on terms a platform can change without notice. Which part of our work still needs a person to answer for it.
CLOSING TAKEAWAY
The quarrel over whether 370 AI-assisted romance novels are literature will not settle anything, because the market has answered in its own way, by rationing shelves rather than judging books. The lesson underneath is harder and more useful. When making something stops being difficult, value moves to whoever controls the method and the shelf, and creators discover that the scarce thing was never production. It was access, and the willingness of a human being to stand behind the result. Both of those are still worth defending.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence's impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan's work at https://www.aiforbusiness.net
Comments