Why Not Just Use ChatGPT to Write a Book?
I use AI every day. ChatGPT and Claude for different jobs, Lex for editing chapters, Grok when I want a contrarian read, and Whisper Flow because typing is slower than speaking. Anyone who tells you a serious publisher must be skeptical of AI hasn't been paying attention. The smartest people I know in publishing use it constantly.
So when a founder asks me why they should pay tens of thousands of dollars for a book service in 2026, when ChatGPT and Claude are sitting right there for $20 a month, I get the question. I'd ask it too.
The answer surprises people. We use AI everywhere in our process at Author.Inc except for the one thing it's worst at, which is writing the book.
The work of making a book is mostly not writing. Most of it is structural, like figuring out the order of chapters so the argument builds and getting a clear picture of what readers retain when they get through a chapter. That work is where AI earns its keep.
When we record an author's manuscript over two days of structured conversation, we tag the transcript as we go. Then, AI then assembles those pieces in the order the outline calls for, and we have a complete first draft within days, with every word coming from the author.
For editing, I lean on Lex.page. It runs on top of the leading language models from OpenAI and Anthropic, but the workflow is dedicated to chapters. You give it a prompt builder up front with your voice, tone, reading level, and patterns to avoid. Then it takes a C-plus chapter through several passes and gets you to a B-plus. From there a human takes over.
For titles, I run candidates through a custom tool that scores them against the qualities of bestselling titles. For reader retention, I feed it a chapter and have it review the work from the perspective of different reader types, including the bullseye reader, the skeptical expert, the dopamine scroller, and the reviewer who is looking for any reason to leave a one-star review.
These are tasks where AI gives a useful directional answer in minutes that would take a human days. I use it constantly. I would be a fool not to.
Generation is where it all falls apart.
A friend of mine, a Harvard grad who was on the founding team of one of Google's AI projects, built a tool that takes a topic and one-shots a full book. It outputs in minutes, and if you compared the result to most self-published nonfiction, it would come out ahead.
It's also completely unusable.
The book has no soul. There is no story you can trace back to a real person, and no insight that could only have come from someone who has actually done the work. The brain registers something is off well before the reader can name what.
This is the problem with the new wave of AI book writing services that promise a finished book in fifteen minutes. The output looks like a book, with a cover and properly formatted chapters. But nobody reads it, because there is nothing in there worth reading.
None of this means AI is useless when it comes to prose. Shane Parrish put two fully AI-generated paragraphs into Clear Thinking as an Easter egg. His editor couldn't identify them, and to my knowledge no reader has correctly guessed which ones they are. They blend in because they're surrounded by his expertise. AI in a manuscript is a seasoning. It cannot be the meal.
Detection is getting better fast.
Anyone who has spent time editing with AI sees the same tells. It loves dashes and antithesis, the rhetorical move of building a sentence around two opposing halves. It writes with a cadence that is even when humans are uneven. Submit a full chapter to a tool like ZeroGPT and a score over 20 percent starts looking dicey. For longer content, the detectors are reliable enough to act on.
Readers don't need a tool to feel it. Reading retention drops in half on content that registers as AI-generated. They don't finish.
There is a related signal from the AI labs themselves. Last year Anthropic settled a lawsuit with authors for $1.5 billion over the books they had used to train their models. The companies building the most advanced AI in the world paid real money for human-written books. They could generate replacement text in seconds if it was viable. They don't, because that text is worth nothing. Original expert authorship is what holds value, and the people building the frontier of this technology have priced exactly what it's worth.
There are two camps in the market right now that both have it wrong.
AI-slop services promise a book in fifteen minutes, producing a book-shaped object that cannot do for a business what a real book does. Traditional ghostwriters and full-service publishers go the other direction, treating AI as a threat to writing as a craft. They charge six figures for a year of back-and-forth and lose the author's voice along the way.
We built Author.Inc around a different approach. The author speaks the manuscript over two days of structured conversation. AI assembles the transcript into a first draft based on the outline we built together. Then we refine the manuscript with the author, with beta readers, and with editing tools that take the draft to a directional A-minus before a human professional editor does the final pass.
Geoff Woods came through this process and published The AI-Driven Leader. His book sold somewhere around a hundred thousand copies in its first year, and the business opportunities the book generated dwarf the copies-sold number. Every word in that book is his. We engineered the architecture around him so he could move at the speed his career required.
Here's Geoff talking about the book he built with us on Robert Breedlove's What Is Money podcast.
It can produce something that looks like a book. The question is whether anyone will read it. When you generate prose without a real expert behind it, you get content with no specific stories and no insight that could only have come from a person who has done the work. Reader retention collapses and reviewers call the book forgettable. For most business purposes the output isn't viable. If your goal is for the book to sell and drive real business opportunities, you need a human expert as the source. AI can help you organize and edit. It cannot be the source.
A ghostwriter interviews you, drafts the book in their interpretation of your voice, and sends it back for review. A traditional ghostwriting engagement runs nine months to a year and often costs six figures. Even with a skilled ghostwriter, the author's voice gets diluted because the writer is a generalist guessing at your expertise. ChatGPT is faster and free, but it has the same fundamental problem amplified. It guesses at your expertise without any of your actual material to draw from. The Author.Inc approach replaces both. The author speaks the content while AI handles organizing and editing, so the voice in the final book is unmistakably the author's.
For short content, detection tools are unreliable. For longer content, like a full chapter or a full book, the detectors are reliable enough to act on. Readers also detect it without a tool. The cadence is too even, and the stories feel generic. Retention numbers drop in half on AI-generated text. Amazon now asks when you upload a book whether you used AI and how much you used it, because they have been dealing with AI-generated books on the platform for years.
Lex.page is dedicated to the writing and editing workflow, with a prompt builder and a chapter-level interface. For research that needs to go deeper than search results, Perplexity or deep-research modes inside Claude and ChatGPT can save you days. Whisper Flow is excellent for voice capture, turning spoken thought into text faster than you can type. Title testing benefits from a custom evaluation against bestseller patterns, which gives a directional read. None of these tools can replace the source material. The source has to be you.
AI-only book services are typically inexpensive, often a few hundred dollars or low thousands, but the output is rarely fit for use in a serious business. Traditional ghostwriting and full-service publishing runs $100,000 or more for an engagement of a year or longer. Author.Inc sits between those two extremes in cost and in timeline. The math works because the author speaks the book, which collapses the time required, and AI handles the organizing work that used to take months of human labor. The voice stays intact because the author is the one speaking.
If you are a founder or CEO running a serious business, and you have been wondering whether the AI tools you already use every day can just write your book for you, I appreciate why you would ask. I use those tools every day too. They are useful and getting better fast. But they cannot be the source of the book itself.
The book has to come from you, in your voice, drawn from work you have actually done. What we have built at Author.Inc is a process that gets that book out of your head and into the world on a timeline that matches the pace of your business.
If you want to see whether this is the right fit, schedule a call with my co-founder Miles Rote.
Copyright ©
2026, Author.Inc. All rights reserved.