The perspiration went missing
Thomas Edison is credited with the line that genius is one percent inspiration and ninety-nine percent perspiration. The Citizen, picking up the story of South African romance writer Coral Hart, turned the formula inside out: with artificial intelligence, the perspiration around creative endeavour can be drastically reduced. Hart, by her own account, has published more than 370 books since February 2025 with AI doing what she calls the “donkey work” of writing. She, as author, still controls the creative process — defining the story concept, a style, details of the characters and chapter outlines — and only then does her system generate the books.
The numbers are the hook, and they are also the first thing to handle with care. Hart reports more than one million copies sold across digital, print and audio, and more than $1 million earned in 2026 so far — about R16.55 million. Those figures are attributed to her; they have not been independently audited. They are also the figures the method she teaches is sold on. One of her novels reached Amazon’s top 100 bestsellers. The vast majority of the titles, most of them romance, are tailored to specific readers on the basis of market research. Hart is a pseudonym. She publishes under more than ten pen names.
How Coral Hart gives every model a job
What makes Hart’s story different from a writer who simply asked a chatbot for a novel is the system itself — and she is the first to say that a bare request produces garbage. “If you say, ChatGPT, write me a book, or, Claude, write me a book, you’re going to get a flaming dumpster fire of trash back because it doesn’t have anything (to work with),” she told AFP. It took her three months to learn the models and build a framework capable of producing a book.
Inside that framework, each model has a job. ChatGPT is “really good at dialogue, terrible at prose.” Claude is “really good story structure, terrible at sex drives.” Grok is “completely unhinged” but good for the “spicy scenes.” The sequence runs like this: Hart sets the concept, the style, the characters and the chapter outlines. The system drafts the whole book in one pass. AI then runs developmental edits, then continuity fixes for dropped threads, flat middles and eye colours that change halfway through. A human proofreader catches the “AI tells” that slip through.
When she started, she said, a paragraph at a time was all the models could manage, taking two or three prompts. Now: “I can now draft the whole book in one sequence.” She is not a newcomer who stumbled onto a trick. Before she started writing with AI around twenty months ago, she had worked in publishing for fourteen years and written several dozen books the old way.
The backlash arrived with the confession
Hart says she received “tens of thousands” of hate messages and several death threats, and is called “grifter” and “villain” on social media. She chose to speak publicly anyway. “I chose to be the one to step up and say, these are conversations we need to start having,” she said. “This is here… They’re not going to switch it off or unplug it because a bunch of authors said so.”
Her answer is commercial, not literary: “Books are selling.” And: “Readers don’t read rubbish books; they return them.” Sales show that buyers wanted the books. What they do not show is what buyers knew about how the books were made — or what they would think if they knew. That is one of the questions the story leaves open. Hart mentions her AI use only when required by a sales platform. “For the last 16 years, nobody’s asked me what my creative process was,” she said.
That is where the argument sharpens. The Citizen’s opinion piece notes that some accuse her of plagiarism, and answers that “all creative art borrows — knowingly or not — from what, and whom, went before.” The accusation is reported, not established, and the borrowing question is genuinely open: whether the training behind Hart’s books involved other writers’ work, and whether that would count as borrowing, is not something the available evidence settles. Nor does the sheer volume settle the quality question. Can 370 books in twenty months each carry real authorial care? The volume raises the question; the reporting does not include an assessment of the books themselves, so whether the output is formulaic remains an open question, not an established fact.
The same argument, from the other side
In a parallel case, Haitian-Canadian novelist Thelyson Orelien’s “C’était ça ou mourir” was withdrawn from the shortlist for France’s Goncourt prize after tests with the AI detector Pangram indicated a 100 percent probability it contained computer-generated text. His publisher, Boréal, said it had confirmed that AI was used in the writing process and ended its professional relationship with him on 21 October 2026. Orelien called the allegations “slander and heresy” and insisted that “not a single sentence in this book was created by a machine, nor the meaning or the structure of a story that comes from my gut.” The book had sold more than 65,000 copies and was due for translation in more than twenty countries.
Hart and Orelien are the same argument from opposite ends. She discloses, on her own terms, and dares the market to judge. He denied the allegations — and his publisher said it had confirmed AI use and ended the relationship. Hart’s own line on the case cuts to the same nerve: “If you need an AI label to hate something, do you hate it, or do you hate AI?” In Orelien’s case, a detector result and a publisher’s statement cost him a prize shortlist and a publishing contract. In Hart’s case, reported sales show commercial demand — and leave open what buyers knew about the process or thought of it.
The books may be the demo
There is a second business behind the books, and it may be the more interesting one. Hart trains several hundred people in her method, and she is preparing to launch software for AI authors called Authorythm. Whether the 370 titles are the business or the proof of concept for the method is an open question — the training and software plans are confirmed, while whether the books are mainly a demo for the method is a question the reporting does not settle. But the shape of it is familiar: a production system, demonstrated at scale, sold as a system.
That is the question Hart’s story actually raises, and it is not whether machines can write. It is which part of writing counts as the author’s work — the writing, or the choosing? She chose the concept, the style, the characters and the outlines. The models did the drafting, the editing and the continuity. A human proofreader did the last pass. If the books sell, the market has answered one question and left the other wide open. Copies are not readers, and sales alone do not show what buyers knew about the process or what they think of AI authorship. But the reported figures show a market, and one author who decided to be the one to step up and say the conversation needs having.
