“South African author Coral Hart says she has used artificial intelligence tools alongside her publishing experience to produce more than 370 books since February 2025 and sell over one million copies across digital, print, and audio formats. Her reported earnings of more than US$1 million in 2026 have drawn attention to the commercial opportunities of AI-assisted publishing, while also intensifying debate about originality, editorial quality, transparency, and the future of human authors.”

How Artificial Intelligence Is Transforming South Africa’s Publishing Industry

A new commercial model for authors

Artificial intelligence is changing the way creative businesses operate, and a South African author’s publishing experience is providing a striking example of how the technology can be incorporated into a commercial writing operation.

Coral Hart, the pen name of a South African romance novelist and publishing entrepreneur, reportedly produced more than 370 books after beginning to incorporate AI into her writing process in February 2025. She estimates that her book sales have exceeded one million copies and that her earnings in 2026 have surpassed US$1 million, equivalent to approximately R16.55 million at the conversion used in the report. These are figures attributed to Hart, rather than independently audited financial results.

The case illustrates a growing shift in the creative economy. AI systems can now assist with story development, drafting, editing, research, and other tasks that previously required substantial amounts of individual labour. For authors who understand their audiences and publishing markets, these tools may make it possible to increase production without expanding a traditional writing team at the same rate.

However, the development raises a more complicated question than whether AI can write a book. It asks how human creativity, machine-generated text, editorial judgement, and commercial demand should work together in a publishing environment where production is becoming faster and cheaper.

From traditional publishing to AI-assisted writing

Hart did not enter publishing without experience. According to reports based on an interview with the author, she had spent approximately 14 years working in publishing and had written numerous books before adopting AI-assisted production.

Her move towards the technology followed a professional setback: she said a major publisher cancelled ghostwriting contracts and replaced her writing with AI. Rather than treating the development solely as a threat, she began experimenting with the tools to understand their capabilities and determine whether she could use them to build a viable business.

That background matters. Her reported approach is not simply to enter a short instruction into a chatbot and publish whatever appears on the screen. Instead, she draws on experience with genre conventions, character development, narrative structure, reader expectations, and commercial positioning.

She reportedly spent around three months becoming familiar with different AI models and developing a repeatable workflow. The experience highlights a distinction between access to an AI tool and the ability to use it effectively. Although the technology can generate text rapidly, a successful publishing operation still requires a clear concept, a defined audience, appropriate editing, and decisions about what readers are likely to purchase.

How AI fits into the writing process

Hart has described using several AI systems for different aspects of her work. Her workflow includes defining the premise of a story, specifying the characters and writing style, and outlining the chapters before asking models to generate a draft.

This preparation provides direction. Large language models generate text based on patterns and instructions, but an open-ended request to write an entire novel may produce inconsistent characters, weak plotting, repetitive language, or scenes that fail to support the wider story.

A structured brief can help establish the boundaries of the narrative and give the model a more coherent starting point. The author can then review the results, identify weaknesses, and request revisions.

According to the reporting, Hart uses further AI-assisted editing to address continuity problems, incomplete character development, dropped plotlines, and weaknesses in the story. Human proofreading remains part of the final process, allowing mistakes and unnatural passages to be corrected before publication.

This approach demonstrates why AI-assisted publishing should not automatically be equated with fully automated publishing. The technology may perform substantial drafting work, but people still need to make decisions about quality, relevance, style, and the final product.

The author’s experience also reflects how quickly generative AI capabilities have developed. Tasks that initially required repeated prompts for individual paragraphs can now be handled through more extensive drafting sequences. Yet producing a longer manuscript in a single sequence does not guarantee that the resulting story will be coherent, original, or ready for readers.

Why the romance market matters

Romance fiction offers a useful example of how a specialised publishing market can support a repeatable production model. Readers often seek particular themes, emotional experiences, relationship dynamics, and familiar genre conventions. Publishers and independent authors can use market research to understand those preferences and position books accordingly.

Hart reportedly tailors many of her books to specific audiences rather than producing manuscripts without a defined readership. She has also published under multiple pen names, separating different parts of her catalogue and protecting her personal identity.

AI may help authors test story concepts, explore alternative plotlines, develop outlines, and revise passages more quickly. These efficiencies can make it easier to maintain a larger catalogue, but they do not eliminate the need to understand readers.

Commercial success depends on more than the volume of manuscripts produced. A book must attract attention, communicate its premise clearly, satisfy its intended audience, and compete with many other titles. Marketing, cover design, pricing, distribution, reviews, and platform visibility can all influence sales.

Hart’s reported results therefore should not be interpreted as proof that anyone can earn millions simply by using a chatbot. Her publishing experience, knowledge of genre fiction, production systems, and understanding of the market are important parts of the wider business model.

The economic opportunity for South African creators

For South Africa’s creative sector, AI-assisted publishing points to possible opportunities for independent authors, editors, translators, illustrators, digital marketers, and small publishing businesses.

A writer working independently may be able to use AI to accelerate preliminary research, organise a manuscript, explore promotional copy, or develop a more efficient revision process. Small teams could also use automation to reduce repetitive administrative work and dedicate more time to editorial decisions, reader engagement, and business development.

The same principle applies beyond books. Similar workflows can support educational materials, online publications, marketing content, and other digital products, provided that creators maintain suitable quality controls and respect applicable legal and contractual requirements.

Nevertheless, the financial benefits are not guaranteed. AI subscriptions, editing, design, advertising, distribution fees, and other operating costs can reduce the money a business ultimately retains. Reported revenue is not necessarily equivalent to profit, and Hart’s reported earnings should not be confused with a verified net-income figure.

South African entrepreneurs considering this model would need to evaluate their own costs, customer demand, platform rules, and ability to produce work that stands out in a crowded marketplace.

A growing debate about creativity and originality

The expansion of AI-generated books has prompted criticism from authors and other creative professionals concerned about the value of human labour, the use of writers’ work in training datasets, and the possibility that publishing platforms will become saturated with low-quality material.

Hart has become a prominent participant in that debate. Reports describe substantial online hostility directed at her, including abusive messages and threats. She has defended her approach and argued that readers ultimately judge books by their quality and whether they find them worth purchasing.

Both the criticism and the commercial argument deserve careful consideration. AI can reduce some production barriers, but greater output does not necessarily create greater cultural value. A marketplace filled with repetitive or poorly edited books may make it harder for readers to discover distinctive work.

At the same time, using AI does not by itself establish that a book is worthless. The important questions include how the technology was used, what the human author contributed, whether the finished work meets readers’ expectations, and whether the process complies with relevant rules.

There are also questions about disclosure. Readers, publishers, and distribution platforms may have different expectations about the use of generative tools. Clear policies can help people understand what is being sold and reduce disputes about authorship and creative responsibility.

Quality control remains essential

AI-generated text can contain factual errors, inconsistencies, clichés, and passages that sound plausible without being well constructed. In fiction, these problems may appear as contradictory character details, implausible motivations, repetitive dialogue, or plot developments that have not been adequately prepared.

For this reason, editorial oversight remains important even when a manuscript passes through multiple AI-assisted revision stages. Human review can evaluate whether a character feels convincing, whether a scene serves the story, whether the emotional arc works, and whether the language matches the author’s intended voice.

A commercially efficient process should therefore be measured not only by the number of books produced but also by the standard of the finished work, reader satisfaction, returns, reviews, and the long-term reputation of the publishing business.

Authors also need to consider copyright, contractual obligations, privacy, and the terms imposed by the services and publishing platforms they use. These requirements can vary by jurisdiction and platform, so creators should check the applicable rules rather than assume that every AI-generated manuscript can be published under identical conditions.

A business built around more than writing

Hart’s wider business activities suggest that the opportunity may extend beyond selling individual books. She founded PlotProse, a training and publishing-tools business, and reports indicate that she has trained several hundred people in her methods. Her operation has also been preparing software called Authorythm for AI-assisted authors.

This model combines content creation with education, workflow development, and publishing support. Such services could appeal to authors who want to learn how to plan manuscripts, organise production, improve editing, and manage the business side of independent publishing.

However, training programmes and software products must demonstrate value in their own right. Prospective customers should assess what is included, how much support is available, what costs are involved, and whether advertised outcomes are realistic. One author’s reported commercial performance is not a guarantee that every student or software user will achieve comparable results.

For South African businesses, the broader lesson is that AI can become part of a service offering when it is combined with domain knowledge, repeatable processes, and a clear understanding of customer needs.

What the story means for the future of publishing

The case of Coral Hart highlights a wider transformation in which the competitive advantage may shift from the ability to produce text alone towards the ability to direct, evaluate, edit, package, and sell it effectively.

As generative systems become more capable, publishers and authors will need to decide which tasks should be automated and which require human judgement. Some will use AI for brainstorming and editing while keeping the writing predominantly human. Others will adopt more extensive machine-assisted production systems, with human involvement focused on planning, revision, and quality assurance.

There is unlikely to be one approach that suits every genre, publisher, or reader. Expectations differ between commercial fiction, literary work, journalism, educational publishing, and specialist nonfiction. The appropriate balance between automation and human authorship will depend partly on the product and the audience it serves.

South Africa’s creative economy can benefit from these developments if creators gain access to useful tools while maintaining standards that protect readers, writers, and businesses. Education, transparent practices, and informed decisions will be important as the industry adapts.

Hart’s reported earnings provide a compelling example of the commercial possibilities of AI-assisted publishing, but they should be understood as an individual case rather than a universal formula. The durable lesson is not simply that AI can generate more books. It is that technology, professional experience, market research, editorial discipline, and business strategy can combine to create a new kind of publishing operation.

Ultimately, the future of AI in publishing will be shaped not only by how quickly machines can produce manuscripts, but also by whether the resulting work earns readers’ trust and delivers a worthwhile experience.