The problem starts before the writing.
From Charlie Hoehn
When I was on Shane Parrish’s podcast, I told him AI sucks as a writer.
Shane pushed back.
“I don’t know if I agree with that.”
He told me he’d put two paragraphs in his book, Clear Thinking, that were completely AI-generated. His editor couldn’t tell. To date, no reader has correctly guessed which ones.
“That’s my point,” I said. “It works for a paragraph or two, or three even. But once it hits a certain point, it stops being good.”
That conversation was in late 2024. AI has improved a lot since then, but almost no books written purely by AI have broken through to the mainstream. Reader retention drops off a cliff, to less than half that of a human-written book. The publishing industry sees these books all the time, and they’re just not very good.
But weak prose is only part of the problem with AI books.
That might sound strange coming from me, considering I co-founded an AI-native writing and publishing company, and we use AI throughout our process. We just never use it to generate an author’s content. I’ll show you exactly where we use it later in this issue.
So why do most AI-written books fail?
The tools that promise to write your book for you are going off minimal inputs. You fill out a form, and out comes a book. The result is embarrassingly bad, because a form can’t tell AI who your book is for, what it should change for them, or why anyone should hear it from you. Everything hinges on clean, excellent inputs. If they’re off by even a little, AI can cost you more time than it saves. And if the prompt isn’t structured, the book turns out poorly.
In this issue, I’ll break down what a good input looks like, and exactly how we use AI to build one.
I spent three years as Tim Ferriss’s Director of Special Projects, and I worked closely with him to launch The 4-Hour Body to #1 on the New York Times list and #1 on Amazon overall.
Before the book, Tim used to hand his protocols to people on a sheet of paper. In Tim’s words, precisely zero of them implemented the advice.
The book had something that sheet of paper didn’t. Early on, Tim tells a story he calls the Harajuku Moment. A friend of his was shopping for clothes with friends in Harajuku, Tokyo. Nothing fit him. Standing outside a boutique, he said out loud that it didn’t matter what he wore, because he wasn’t going to look good anyway.
Hearing himself say it, he realized he’d spent over a decade making excuses about his health. He was successful in every other part of his life except taking care of his own body. Enough was enough.
Readers saw themselves in that moment, and it hit them hard.
Tim also made himself the human guinea pig. When he put on muscle, he showed the photos and laid out exactly what he did to get those results.
Same information. Completely different results.
The book changed thousands of lives, including mine, because of the stories, the sequencing, the designed path, the unique positioning, and the relationship and trust Tim built over hundreds of pages. That path and that trust were always what readers were paying for. The information was the least of it.
That’s exactly the part AI can’t produce on its own. When left alone, AI produces average, predictable work. And information alone is now free everywhere. The market is repricing it at its new value, which is zero.
You can see how this plays out for bestselling authors right now, like Mel Robbins and Alex Hormozi. Their readers could get similar advice anywhere. What they’re buying is a relationship with someone they already trust, and the information comes along with it.
The author is now part of the product.
A chatbot doesn’t have a real face or story, and it will never be a personality anyone roots for.
It reminds me of Harry Harlow’s experiments with baby rhesus monkeys in the 1950s. He gave them two stand-in mothers. One was bare wire and held a bottle of milk. The other had no food, but it was covered in soft cloth. The babies went to the wire mother to feed, then spent most of their time clinging to the cloth one. They needed the milk to live, and they still chose warmth and comfort.
I think something similar is happening with books. We don’t open a book purely for information anymore. The internet solved that problem a long time ago, and AI distills information even better. What readers are looking for now is warmth, connection and humanity, especially when so many people feel disconnected.
So the job of the book has changed. It used to be a place people went for information. Now it needs to connect the reader to the right human being on the other end: an expert they know, like and trust.
That’s why I’m excited about where books go from here. The next generation of nonfiction will have to earn attention through story, judgment, trust and transformation, and AI will give more people than ever a chance to create those books.
If you’re curious, read the response I wrote to Tim’s article on AI and how-to books.
Miles and I started Author.Inc because publishing wasn’t built for experts. The traditional route can take years, and you often give up your rights along the way. Ghostwriters can hand back pages that sound like someone else. And writing it yourself means months at a blank page while your business needs you somewhere else.
For decades, that grind filtered out everyone who wasn’t a professional writer, no matter how much they knew. AI removed the filter. The expert with 20 years of hard-won answers and zero writing chops finally has a shot at creating a book.
Chris Perkin, author of Lead & Let Lead, speaking his book.
Very rarely does anyone get talker’s block, but they do get writer’s block all the time. So, our authors speak their books instead of typing them. AI handles the slow work around those recordings, which shrinks the traditional publishing timeline to a first draft in 30 to 60 days and a published book in 6 to 8 months. You keep the IP, the royalties and the rights. There is no AI generating content. The book sounds like you, because every page has your own words.
This approach to AI has gotten attention outside publishing, too. Zapier named me its 2025 AI Solution Innovator of the Year.
If you give AI a clean, structured input, it can do some pretty cool stuff. That’s why a plan needs to come first. Before anyone writes a chapter, we get clear on four things, and then AI helps turn the author’s own words into a book.
One specific person you know by name. You should be able to describe what they were doing yesterday, and the moment last week when they felt the problem your book solves. This step stays human. No tool can pick your reader for you.
The specific change your reader wants, in their health, their wealth and their relationships. I built an AI-assisted Blueprint tool that walks the author through these questions one at a time and pushes past the first, shallow answer. The author answers every question, and an editor reviews each one.
How the book makes money for you beyond royalties: consulting, speaking, advisory work, a mastermind. The book should activate at least one of them. The Blueprint tool covers this too.
AI suggests structures from the Blueprint, drawn from the shapes strong nonfiction books use. The author chooses and shapes the chapters, then tests them with their own readers before recording.
The author speaks each chapter from talking points. AI transcribes those recordings and organizes the author’s own words into the chapter structure we planned, with guardrails that keep it from changing their wording. It tightens at the sentence level and nothing more. Then human editors take the manuscript from there. When an author revises, AI can make the edits they ask for, but it never does the editing for them. The judgment stays with the author.
You can start on the first four today.
Copy the prompt below into ChatGPT or Claude:
One more tip: if you want to go further, ask the AI to write the ideal prompt for the outcome you want first. AI is better at communicating with itself than you are.
Geoff Woods wrote The AI-Driven Leader, a book about using AI to make faster, smarter decisions.
Before he wrote a word, he took his table of contents to more than 200 CEOs. Geoff was one of the first authors through an early version of our process, back in 2024. He taught each chapter out loud to a live person, and we used AI to turn those transcripts into his draft, with prompts that kept it strictly on guardrails: don’t mess with his wording.
From day one to a complete manuscript took 39 days. The book was fully published in five months, and it sold over 100,000 copies in its first year. He started with roughly 2,000 LinkedIn followers.
Of course, results like Geoff’s aren’t a guarantee, and they aren’t typical. But they show what’s possible when the plan comes first and AI does only the work it’s good at.
You share your idea, your audience and what you want the book to make possible. Then you have a strategy session with an Author.Inc editor, and our team reviews the idea through four lenses: the reader, the promise, your authority and the business case.
The guided Blueprint tool I built helps you get your answers down, and our editors review every one. In 7 to 10 business days, you get a custom Blueprint laying out your book section by section. This covers the big idea, the reader, the problem, the business case, the market and the chapters in order. If you go on to write your book with us, it credits toward our Publish program.
See the Book BlueprintIf you’d rather talk through your idea first, my co-founder Miles would love to hear it. Book a call with Miles.
And if you know someone with a book in them, feel free to send this along.
Copyright ©
2026, Author.Inc. All rights reserved.