More and more clients come to me with a book they tried to write with AI. It came out badly, they know it came out badly, and they want it fixed. I feel for them, because most of them spent months on that draft before admitting it wasn’t working.
AI used as a tool works fine. Hand it the writing, and the manuscript reads like a brochure and ends up on my calendar. If that’s your book, you’ve got a question to answer before you publish: when Amazon asks whether AI was used, what do you say?
Amazon’s written rules answer it, and they’re shorter than the forum threads about them. I don’t trust Amazon to be reasonable with authors, so I read those rules literally, and you should too.
Think of the AI question as a customs form. The officer wants to know what’s inside your suitcase and where it was made. Goods made by a machine go on the form, even if you repainted them before the trip, and a human ghostwriter who packed the bag for you never does.
Do I have to tell Amazon my book used AI?
Only if AI generated the content. Amazon’s rule sits in the KDP content guidelines. It reads: “We require you to inform us of AI-generated content (text, images, or translations) when you publish a new book or make edits to and republish an existing book through KDP.” The next sentence matters just as much: “You are not required to disclose AI-assisted content.”
Your answer depends on which side of Amazon’s line your book falls. A spell checker, a grammar tool or a chatbot you used to brainstorm doesn’t trigger the disclosure. Nobody declares a spell checker at customs. Text that a chatbot wrote does trigger it, and so does a cover image an AI tool drew.
Translations count too. If you had an AI tool translate your ghostwritten memoir into Spanish or German, the translated edition is AI-generated under Amazon’s definition, even though a human wrote the original. Declare the translation, and have a fluent human read it before a single copy sells.
Watch the word “republish.” If you revise an older book and upload it again, the question comes back, and your answer has to reflect the version you’re uploading now.
What’s the difference between AI-generated and AI-assisted content on KDP?
Amazon draws the line at who created the words. In its own definition, AI-generated content is “text, images, or translations created by an AI-based tool,” and it adds that content counts as AI-generated “even if you applied substantial edits afterwards.”
AI-assisted content is the opposite case. Amazon says that if you created the content yourself and used AI tools “to edit, refine, error-check, or otherwise improve that content,” it’s AI-assisted. It also says that using an AI tool to brainstorm and generate ideas, where you “ultimately created the text or images yourself,” is AI-assisted.
The “substantial edits” clause catches people. They assume that rewriting every sentence of a ChatGPT draft turns it into their own work, and under Amazon’s definition it doesn’t.
Nobody at customs asks how hard the repainting was, either. People who spent months polishing an AI draft hate hearing this, and I understand why it feels unfair. A wrong answer on the form causes far more trouble than a yes you didn’t strictly need, so when the draft started in a chatbot, tick the box.
Does Amazon’s AI question apply to a ghostwritten book?
Only to whatever AI tools touched it. KDP asks about content created by an AI-based tool, and a ghostwriter is a person. Someone other than you writing the book doesn’t make it AI-generated.
Amazon does care that you have the right to publish it. KDP’s intellectual property FAQ says the terms and content guidelines “require that you hold the publishing rights to any content you upload,” and it tells people publishing content someone else created to gather their documentation before submission in case KDP asks. For a ghostwritten book, that documentation is your contract and its copyright assignment.
Before you fill in the form, find out what your ghostwriter did with AI. If they used AI to generate chapters and then edited them, Amazon’s definition makes those chapters AI-generated, and you’re the one clicking the button. My article on whether a ghostwriter will secretly use AI on your book covers how to ask, and what the contract should say.
For the record, I use AI every day as a digital assistant: transcripts of interviews, summaries of each session that I send to the client, first-pass outlines. I write the books. Under Amazon’s definitions, transcripts and idea work sit on the assisted side, because a human created the text in the book.
Recordings help here too. I record every interview, say so up front and put it in the contract. Clients can ask for copies. The contract keeps those recordings for three months after the project ends, so request yours before then if you want proof that a person built your book from your own stories.
What if my ghostwriter fixed a book I first wrote with ChatGPT?
It depends on what survived. If sentences the AI created are still in the book, edited or not, Amazon’s definition makes that content AI-generated and you should disclose it. When the ghostwriter threw the draft away and wrote the book fresh from interviews, using the AI version only as a map of what you wanted to say, a human created the text.
Most rescue jobs fall somewhere in the middle, and the line there gets blurry. When you can’t say with confidence that a person wrote every sentence, answer yes, and ask your ghostwriter to tell you in writing which passages came through from the old draft.
Rescue work costs what it costs because the AI draft rarely gives a ghostwriter much to keep. My piece on what it costs to fix an AI-written manuscript explains why, and it’s the same reason the disclosure question gets easier after a real rewrite.
Can I sell a book written by ChatGPT on Amazon?
You can. KDP allows AI-generated books as long as you disclose them and they meet the content guidelines, and Amazon says publishers are responsible for making sure all AI-generated and AI-assisted content complies, including with intellectual property rights.
Ownership is harder. The U.S. Copyright Office’s registration guidance on works containing AI-generated material says copyright “can protect only material that is the product of human creativity.” Applicants have to disclose AI-generated content that’s more than de minimis and exclude it from the claim. Text the machine wrote gets no copyright protection, so you can’t stop others from copying it. My article on AI copyright for writers goes deeper on that.
Quality is the other trap. Amazon’s content guidelines also cover books that disappoint customers, and Amazon reserves the right to remove content from sale if it determines the content creates a poor customer experience. Its storefront is already choking on machine-written junk published by content mills chasing keywords. I hate what those operations have done to the category pages, and a platform that tolerated them for years now has every reason to be quick with the removal button.
And Amazon isn’t the whole market. IngramSpark’s catalog integrity guidelines, published in October 2023, list “content created using automated means, including but not limited to content generated using artificial intelligence or mass-produced processes” among the things it won’t carry. So IngramSpark, the distributor I use for bookstores and libraries, can refuse a book whose text a machine generated.
What happens if I answer Amazon’s AI question wrong?
Amazon says it will reject or remove content that doesn’t follow its guidelines. When it removes a title, it notifies the publisher, who can appeal. That’s the written rule, and how it plays out is up to Amazon.
I don’t put much faith in that appeal. My own KDP account was terminated after I followed Amazon’s instructions to change a subtitle, a story I tell in why Amazon KDP terminated my account. When a platform can do that to an author who followed the rules, don’t hand it a reason by answering a disclosure question carelessly. Have a lawyer read your ghostwriting contract if the ownership or AI terms are unclear, and answer the form truthfully.
A wrong answer can also come back long after launch. If you told Amazon no and the book later draws complaints that it reads machine-written, you’ll have to defend what you declared. Keep the ghostwriter’s written confirmation, the interview recordings and the session summaries together in one folder so you can. It makes me angry that careful authors have to build a paper trail like this when the mills that caused the problem never kept one.
What should a ghostwriting contract say about AI?
It should say whether the ghostwriter is allowed to use AI to generate any of the text, and if so, require them to tell you in writing which parts. Put it in the contract before the first interview. A promise made on a sales call won’t help you when you’re staring at Amazon’s form a year later.
The clause I’d want as a client is simple: the ghostwriter writes the book, AI tools can handle transcripts, summaries and other assistant work, and none of their output goes into the manuscript as text. At delivery, the ghostwriter confirms in writing that a person wrote the manuscript. With that in hand, you can answer Amazon’s question in ten seconds and sleep fine.
Agencies are where this goes wrong. You’re sold by a salesperson and handed to writers you never met, and nobody in that chain has promised you anything about how the chapters got made. My article on AI ghostwriting vs human ghostwriting shows what the interview process produces that a chatbot can’t.
Some agencies will tell you their writers never touch AI and then refuse to put that sentence in the contract. Believe the refusal. A salesperson’s promise binds nobody who writes your chapters, and when the form asks the question a year later, you’ll be the one answering it with nothing in writing behind you. Ask for the clause and the delivery confirmation, and if either one gets dodged, walk. It infuriates me that people pay agency prices and still can’t get a straight commitment about who wrote their book.
Have a lawyer read the whole ghostwriting contract before you sign, with the copyright assignment and the AI clause side by side. Together they’re your answer to both questions Amazon can ask: do you own this, and who made it?
Is disclosing AI to Amazon the same as telling readers?
No. Amazon’s form tells Amazon. Readers and reviewers are a different audience with different expectations, and whether you owe them a line in the acknowledgments is a separate decision. My article on AI disclosure and what you owe readers covers that side. Plenty of authors thank their ghostwriter by name in the acknowledgments. A thank-you like that has no bearing on your answer to Amazon’s form, since the form asks about software and a ghostwriter is a person you hired.
For a ghostwritten book, the AI question and the ghostwriter question also stay separate. Whether you tell anyone a ghostwriter wrote your book is your call. AI is a different matter, because Amazon has asked about it in writing. For the book itself, the tool matters less than whether a reader stays engaged, and a machine’s first draft with your name on it is still a bad book whichever box you tick.
Declare what’s in the suitcase
Before you upload, ask your ghostwriter in writing whether any AI-created text is in the manuscript, and keep the answer with your contract. If you wrote or co-wrote parts with a chatbot, go through the file and mark them. Then fill in Amazon’s form to match what’s in the book, and if you want the book in bookstores and libraries through IngramSpark too, check whether a yes on Amazon’s form closes that door.
The content mills that flooded Amazon made this question necessary, and careful authors are the ones filling in the form because of them. That makes me furious. Don’t let it make you careless. If you want a book written by a person from your own stories, my ghostwriting services are built that way, and the Book Discovery Intensive is a shorter first step. The Ghostwriting Hub has more on AI and ghostwriting.
Pack the bag by hand, and you can walk through customs with nothing to declare.
