You Used AI and It Shows. Here’s How to Fix It.

This entry is part 30 of 29 in the series Artificial Intelligence for Writers
TL;DR: You cannot rescue an AI-written book by sprinkling humanity on top of it. Swapping out robotic phrases and working in a personal story does not fix it, because what makes machine prose dead is in the bones, not on the surface. If you wrote your book with AI and suspect it shows, you are probably right, and it is fixable. Just not the way almost everyone tries.

You cannot rescue an AI-written book by sprinkling humanity on top of it. People try constantly. They swap out the robotic phrases, work in a personal story, change “in today’s fast-paced world” to something a person might actually say, and assume the thing now passes. It does not. What makes machine prose dead does not sit on the surface where the fixing happens. It is in the bones, and you cannot sand it off how I fix AI prose on a book.

If you wrote your book with AI and some part of you suspects it shows, you are probably right, and it is fixable. Just not the way almost everyone tries to fix it.

Why the Sprinkle Never Works

The mistake is treating AI prose as a word problem. For more, see 61% of writers use ai. here is what that means. People think if they hunt down the tired phrasing and dust the chapter with a little warmth, the deadness lifts. It does not, because the deadness is structural. For more, see mit research shows ChatGPT weakens your brain — a profession. The machine built the whole chapter the way it builds everything, hitting the expected beats in the expected order, saying the average of what everyone says about the subject. It opens the same way every chapter opens, runs the same length, makes the same shape of argument, and closes with the same tidy bow. Changing the paint does not move the walls, and the walls are the problem.

I watched a client learn this the hard way. He sent me big chunks of his manuscript that he had written with AI and asked me to take it from there, and it was obviously AI. Anyone who reads for a living sees it in a paragraph, the way you hear a backing track under a singer who is half a beat off. I showed him the specific tells and how that prose would quietly undercut everything he wanted the book to do. He believed me. Then he went back and tried to save the AI chapters himself, working in some warmth, a story here, a joke there. The result was better than raw machine output and still not good. It read like a hollow thing wearing a couple of human accessories, because that is exactly what it was.

What gives it away is not one thing. It is a pattern. The writing hedges everything and commits to nothing, because the machine has no stake in any of it. It states the obvious with great confidence and never says the one specific, slightly risky thing that only a person who was actually there would know to say. It reaches for the same handful of transitions and the same shape of sentence until the rhythm flattens into a hum. And it is relentlessly even, every paragraph weighted the same, no moment that matters more than the others, because nothing matters to a machine. Put those together and you get prose that is technically fine and emotionally vacant, and a reader feels the vacancy long before they could explain it.

The only thing that actually worked was rebuilding the chapters from his real material, in his real voice, which is rewriting, not editing. That means going back to what he actually knew, the experiences only he had, the way he actually talks, and building the chapter out of those instead of patching the machine’s version. It is more work than people want it to be. It is also the only thing that works, because you are not cleaning up a draft, you are finally writing the one the machine could never write, the one with a person inside it.

Stop Worrying About Getting Caught. Worry About Whether It’s Good.

If you spend any time in writing forums or the Facebook groups, you know the fight, and it runs close to an even split. One camp treats anyone who touches AI as a sellout bound for hell, loud and certain. The other camp, quietly the bigger one, keeps its mouth shut and uses AI however fits their workflow, because they have books to finish. Both camps are arguing about the wrong thing, and so are you if you are lying awake worried someone will find out. Readers do not run manuscripts through detectors before deciding whether they like a book. A bad book gives itself away on its own, and a good one rarely draws the scrutiny. For the full case on why the tool was never the real problem, only what it produces when you let it run the show, I made the argument in a companion piece.

Here is my honest position, and it is more specific than either camp wants to deal with. I use AI where it is genuinely good and does no harm. Organizing dozens of interviews into an outline, building summaries, handling communications, the machine is excellent at that and I lean on it. I run grammar checkers too. What I never touch is the replacement text they suggest, because that is exactly where the AI phrasing sneaks in, one harmless-looking rewrite at a time, until the page sounds like everyone and no one. And when I write the actual book, every word is mine. Not a sentence of a client’s manuscript is written by a machine.

That is not purity for its own sake. It is the only line that produces a book worth reading, and it draws a clean border between the work AI does well and the work it ruins. Organizing is logistics, and the machine is a fine clerk. Writing is judgment and voice and lived specifics, and there the machine has nothing to offer because it has nothing to draw on. It was never in the room. It never made the bad call, never lost the account, never sat across from the person whose story this is. So it guesses, and a guess is exactly what a reader feels and exactly what a reader puts down.

The question, in the end, is never which tools touched the book. The question is whether it is good. Is it good. Does it do what it set out to do. Is it different from the hundred other books on the same shelf. Does it matter to the reader who picks it up. Those are the only things worth asking, and a hollow book fails every one of them no matter how it was made. So stop asking whether anyone can tell. Build it back from your own voice, and the question of how it was made stops mattering, because the only thing a reader can feel on the page is whether a real human being is home.

If you are sitting on an AI draft right now, the practical path is short. You have three options and only one of them works. You can keep patching it and ship something hollow, which is the road most people take and the road that wastes the year. You can rebuild it yourself from your real material, which works if you can actually write and have the time. Or you can hand the raw knowledge, the interviews, the notes, the stories, to someone whose entire job is turning that into a book that sounds like you. What you cannot do is sprinkle your way out of it. The draft is not the problem to solve. The draft is the evidence that the real book, the one only you can supply the inside of, has not been written yet.

Schedule a free consultation to discuss your book project.

Frequently Asked Questions

Can you fix an AI-written book by editing it?
Not by surface editing. Swapping out robotic phrases and adding a few anecdotes does not work, because what makes AI prose hollow is structural, not cosmetic. The machine built the whole chapter the way it builds everything, hitting average beats in the expected order. The only fix that works is rebuilding from your actual material in your actual voice, which is rewriting, not editing.
Should I worry about getting caught using AI to write my book?
The caught question is a distraction from the real one. Readers do not run manuscripts through detectors before deciding whether they like a book. A bad book gives itself away on its own, and a good book rarely draws the scrutiny. The thing to worry about is not whether anyone can tell. It is whether the book is actually good.
Is it ever okay to use AI when writing a book?
As a tool, yes, where it does no harm: organizing interviews into an outline, building summaries, handling communications. Grammar checkers are fine too, as long as you ignore their replacement text, which is where AI phrasing creeps in. The hard line is the prose itself. The actual writing of the book should be entirely human, because that is the only thing that produces a book worth reading. The test in the end is whether the book is good, different, and does what it set out to do.

📝 Disclaimer

The views and opinions expressed in this blog post are solely those of Richard Lowe and are based on personal experience and research. This content is for informational purposes only and should not be construed as professional legal, financial, accounting, or business advice. Always consult with qualified professionals before making important business or legal decisions. Richard Lowe is not a lawyer, accountant, or licensed professional advisor, and this content does not establish any professional relationship.

Leave a Reply

Your email address will not be published. Required fields are marked *