Self Publishing With AI: The Honest Money Math

Self publishing with AI will not make you rich, and I'd rather lead with that than sell you the dream for the length of an article. Most Kindle books earn close to nothing. Mine have earned close to nothing in plenty of months, and I run an AI-assisted fiction series on Amazon where I make every story call and read every chapter before it ships.

Quick answer if you're skimming. Yes, you can make money self publishing books with AI help. A small slice of authors do genuinely well, most make coffee money, and the amount you keep per copy is smaller than the ads imply once you subtract the fee nobody screenshots. Amazon pays most indie ebooks either 70% or 35% of the list price, and I'll run that whole calculation below with real numbers. The one reframe I'd hand you before anything else is this. Kindle royalties are a lifetime total, not a monthly salary, and almost every guide skips right past that part.

What Self Publishing With AI Actually Means

There's a fantasy version of this where you type a prompt, a finished book falls out, you upload it, and money arrives while you sleep. That book exists. It's on Amazon right now by the thousands, and readers are one-starring it into the ground.

The real version is narrower. AI does chunks of the work, drafting, outlining, edit passes, cover concepts, formatting, and a human decides what the book is and whether each part is any good. That's the whole game. When I checked what currently ranks for this search, the top results were a half-finished Medium listicle, a Reddit thread complaining about low-effort AI books, and a couple of tool landing pages. Not one of them showed the money math, which is strange, because the money math is the only thing that tells you whether any of this is worth your weekend.

So that's what I'm going to do. The rest is receipts.

The Royalty Math Nobody Screenshots

Here's the part the course ads leave out. On Amazon, most ebooks earn either the 70% or the 35% royalty rate, and the 70% rate only applies if you price inside a specific window. The floor is $2.99. Below it you drop to 35%, no exceptions. And the 70% tier also charges you a delivery cost, 15 cents per megabyte on Amazon.com, which comes off the top before your cut. The 35% tier skips the delivery fee but pays half as much, so it almost never wins.

The formula Amazon uses is 70% times (list price minus tax minus delivery). A typical text ebook is around a megabyte, so I ran the numbers at a flat 15-cent delivery. Image-heavy books cost more to deliver, plain novels often cost less.

Sit with the $4.99 row for a second, because that's the price most indie fiction lands on. You keep about $3.39 a copy. To clear a thousand dollars from one title you need to sell roughly 295 copies, and 295 sales of a brand-new book by an unknown author is a genuinely good month, not a floor. This is why nobody serious pins their income to a single book.

The $2.99 Cliff, and the New Ceiling

Watch the bottom of that window. Price your book at $2.98 and you don't earn one cent less than at $2.99, you earn less than half, because you've fallen off the 70% tier onto 35%. A $2.99 book pays you $1.99. The same book at $2.98 pays about $1.04. One cent of list price, nearly a dollar of royalty, gone.

There's an upside change worth knowing too. The top of that 70% window used to be $9.99, and Amazon raised it to $12.99 as of July 7, 2026. So the ceiling on what you keep per copy moved from about $6.89 up to roughly $8.99, if your book can actually command that price. Most AI-assisted first books can't, and pricing above what readers think you're worth is its own quiet way to sell nothing. Still, the room is there now, and longer nonfiction is where I'd test it.

Where AI Earns Its Keep, and Where It Makes Slop

I'll be specific about the parts of my own pipeline that pull their weight.

Drafting and edit passes, huge. Work that used to eat months now takes weeks, which is the only reason a one-person catalog is even possible. Cover concepts, also great, and cheap in a way that surprises people. I generate cover directions on my own machine, so after the initial setup each image costs me basically electricity, something like fifty seconds a render. Formatting the manuscript into a clean Kindle file, scripted once and reused. None of that is glamorous and all of it saves real hours.

Now the part the tool ads won't tell you. The single fastest way to lose money here is to let the model be the author. If you hand readers a book the AI wrote start to finish, it reads like one, and they can tell faster than any detector. They leave the reviews. A couple of brutal one-stars early and the listing is effectively dead, because Amazon's ranking buries it and no amount of ads revives a book with a 2.8 average. I've watched this happen to other people's titles and it is not subtle.

Amazon also asks you to disclose AI-generated content when you upload, and they've tightened how many new titles a single account can push out per day, both of them aimed squarely at the spam-farm version of this. The quality bar isn't a suggestion. It's the whole business. My pipeline only works because I make the story decisions and read every chapter before it goes live, which sounds obvious and is the exact step most people skip to scale faster.

Lifetime, Not Monthly

I keep hammering this because it's where the disappointment lives.

When someone says their book "makes $300 a month," they usually mean it made $300 once, at launch, and now makes eight. Kindle income front-loads hard and then trickles. One indie author documented publishing a book almost entirely through an AI workflow and sold 512 copies in launch week for roughly £330, about $445, and they were honest that they'd priced it at 99p on purpose to chase reach over royalties. Their own line about it was that a number one bestseller at 99p does not make you rich, and it was never supposed to. That's the healthiest framing I've read on any of these pages.

My own honest number is fuzzy on purpose. Books are one slice of roughly $10,000 in lifetime revenue across everything I've shipped, seven toolkits and the fiction included, and I'm not going to carve out a clean per-book figure because the real one wobbles month to month and any tidy number I gave you would be half marketing. What I will say is that the boring truth of self publishing, AI or not, is catalog and distribution. I once ran a content site that sat at Google position 47 with four visitors a day, and the lesson transfers directly. Making the thing is maybe a third of the job. The rest is getting anyone to find it. Books built with AI don't escape that, they just let you produce the catalog faster, which only helps if the catalog is good. I get into the wider version of this in the pillar on how to make money with AI, where self-publishing ranks squarely in the middle of nine options.

Straight Answers to the Questions People Actually Search

Can I legally publish a book written by AI? Publishing it, yes. Owning it, complicated. The US Copyright Office won't grant copyright to purely AI-generated work because it requires human authorship, so the more of the creative direction is genuinely yours, the more protected you are. Practically, make the real decisions yourself and keep your notes.

Can I sell a book I wrote with ChatGPT? Yes. Amazon allows AI-assisted books and just asks you to disclose AI-generated content at upload. The catch is the same quality wall as everything above, a book that reads like raw ChatGPT output gets treated like raw ChatGPT output. If you want the workflow side of drafting with an assistant, I wrote that up separately in make money with ChatGPT.

How many books do I need to sell to make $100,000? Off that royalty table, a $4.99 book at about $3.39 a copy means roughly 29,500 copies. From one title that's a career-defining bestseller. Spread across a catalog of a dozen titles over a few years it's a real, if demanding, small business. Anyone quoting you a faster version is selling the course, not the outcome.

Can Amazon tell if a book was written by AI? Detection is unreliable and honestly a little beside the point. The disclosure question and the daily upload limits are the guardrails Amazon actually leans on, and the real judge is your reviews. Readers are a far better AI detector than any software, and they're unforgiving.

Can I publish a book illustrated by AI? Same rules as the text. Disclose the AI-generated images, make sure you control the rights, hit the quality bar. A children's book with obviously mangled AI art gets torched in reviews just as fast as a lazy novel does.

So, Should You Do It

If you're picturing passive income by next month, no, and I'd rather you hear it from me than from a refund request. If you like the actual craft, you're willing to treat AI as a fast assistant instead of a ghostwriter, and you can stomach a few zero months while a catalog builds, then it's a real and legitimate thing to do. That's the version I run.

The fastest money in this whole space isn't books anyway, it's services and products with quicker paydays, and I laid those out in the guide on selling digital products with AI. Books are the slow crop. Worth planting, in my experience, just not the harvest you eat this season.