The question went up on a Monday night in a Facebook group for people self-publishing on Amazon. Why do people care so much if somebody uses AI? You didn’t put the effort in, you’re not selling it, you have no stake in it. The poster said she’d watched people get ripped apart over it and wanted to understand why.
Answers came fast: AI is trained on stolen work, AI is slop, AI is boiling the planet, you’re jealous, you’re a con artist, you share one brain cell with the other pro-AI zombies, you’re not an author but a prompt enterer. Somebody called the question disingenuous. Somebody else said the fight had become like religion and politics, where everyone has to believe what you believe or they’re wrong, and the reply to that one opened a fresh argument about whether she’d missed the point.
I read every one of them, and my reaction was mixed. That surprised me. The thread is a mess and parts of it are ugly. It’s also the best collection of real answers to that question I’ve seen in one place. You have to dig for them.
The stall next to yours
Picture a Saturday farmers market. Forty stalls, every vendor paying the same fee, every one of them hauling in tomatoes and honey and bread they grew or baked themselves. Then a new vendor sets up at the end of the row with a truck full of factory tomatoes, sells them at half price under a hand-lettered sign that says “homegrown,” and brings in a fresh truck every morning.
Somebody asks the other vendors why they care. It’s his stall, and he’s not selling their tomatoes.
The honest vendors have good answers. The market has a limited number of stalls and a limited number of shoppers, and he’s taking both. The sign is a lie, and the lie makes shoppers suspicious of every sign in the market. And if some of those tomatoes were stolen from their own fields, they’d like a word about that too.
They also have bad answers, and some of them get a little carried away. A few start shouting at every vendor with a nice-looking tomato, including the woman who’s grown them on the same plot for thirty years. That’s the thread in one picture.
Why do writers care if other authors use AI to write books?
Because a self-published author shares the store, the payment pool and the reader’s trust with everybody else who publishes there. A book written with AI competes for all three. The question in that thread assumed one author’s AI use touches nobody else. For fiction on Amazon, that assumption is wrong.
The clearest answer in the thread came from a romance author who has written full time for ten years. She said that for every human-written book in her part of Amazon there are now a hundred AI books, many of them released in batches by companies running ads. Those books go into Kindle Unlimited, and she has to share her income with them. She’s scared she won’t be able to pay her bills.
That last part stuck with me. She’s describing her mortgage. What would you call that, if not a stake?
Kindle Unlimited pays authors out of a single monthly fund, divided by pages read. Every page a subscriber reads in a mass-produced AI book is a page that didn’t get read in hers, and the money follows the pages.
If you publish in Kindle Unlimited, you’re in that pool too, whether you’ve ever thought about it or not. Amazon itself saw enough of a flood that it capped the number of new titles an account can publish in a single day back in 2023. A cap like that doesn’t get invented for a problem that doesn’t exist.
Another commenter said agents are drowning in machine-written submissions and can spare a few minutes for each one. I can’t verify his numbers, and I’m not going to repeat them as fact. The shape of the complaint is familiar to anyone who has watched submissions pile up: more volume, and less time for each real manuscript.
So the romance author has a stake. So does every new writer trying to get noticed in a category that fills up with a hundred look-alike covers a week. When Bob publishes a book written with AI, Sally cares because Bob’s book is standing between Sally and her readers.
Were authors’ books used to train AI without permission?
Yes, and in at least one major case a court and a settlement put numbers on it.
In the thread, a novelist said sixty of her books had been scraped without her permission. Another commenter demanded proof, and she didn’t have the link handy because she was at an airport. The exchange got nasty. Then a third author stepped in with the actual history, and she was right.
Bartz v. Anthropic was a suit by authors over books the company downloaded from pirate libraries, LibGen and the Pirate Library Mirror, to build a training collection.
The judge in that case ruled that training on books acquired legally was fair use and that the pirated library was a different matter. Anthropic settled for $1.5 billion, covering roughly 500,000 titles, about $3,000 per work before fees. According to the Authors Guild’s summary, eligibility depended on the book being registered with the Copyright Office in time. The author in the thread pointed to copyright certificates for that reason. She said 22 of her books were on the list. Would you shrug off AI if 22 of your books were on it?
So “AI stole my books” is a claim with evidence behind it for a lot of authors. It’s also narrower than the way it gets used. The ruling drew a line between books bought and books pirated. “Everything AI produces is theft” is a moral position some people hold, and they’re entitled to it, but it isn’t what the court found.
One commenter in the thread tried the opposite overstatement: that AI is mostly trained on public domain material. That’s wrong too. The Bartz case exists because it wasn’t.
If your own books were registered in time and you missed the claims window, that’s a hard lesson in why registration matters. I wrote about AI and copyright for authors separately, because the rules keep moving.
The label on the jar
The quietest comments in the thread made the strongest point, and almost nobody argued with them.
One man said he had nothing against writers who admit they use AI. He had two concerns: people who use it and lie about it, and readers who have no idea what they’re reading. A woman said anything made with AI should be labeled so people can decide for themselves. Another said there’s no tagging system that tells a reader what’s human-made, so readers trying to support human authors can’t.
That’s a reasonable request. Kindle Direct Publishing asks authors during setup whether AI generated their text, images or translations. Amazon collects that answer and doesn’t show it to the people buying the book. The disclosure goes to the store and stops there. Why isn’t it on the product page? Amazon already has it.
Would you want to find out, halfway through a novel you paid for, that nobody wrote it? A reader who wants to avoid AI-generated fiction has no way to do it on Amazon except by guessing, and guessing is exactly how the witch hunts start.
I’ve written about what authors owe readers when AI touches a book, and the short version applies here: tell the truth where it’s asked for, and don’t pass off generated work as your own. A label would settle a lot of fights in threads like this one before they started, and I’d welcome it on every book I publish.
Which arguments about AI in writing don’t hold up?
Plenty of them, and they came from both sides of the thread.
One commenter blamed fourteen weeks of heat over 95 degrees on AI. Data centers use a lot of electricity and water, and communities near them have real grievances about noise, power bills and water use. A regional heat wave is weather, and pinning it on the thing you already dislike is how an argument loses the people it needs to persuade. Who’s going to trust the rest of your case after that?
Several people called anyone who touches AI a thief, including a man who described his process in detail. He works out the plot and every action beat himself, talks through ideas by voice while he drives, and uses AI to clean up grammar and commas he’s never been good at. A commenter told him he wasn’t an author and was stealing from real authors. Another told him he was stealing from editors. He answered with sarcasm, and the conversation ended there.
You can argue about where the line sits. When I sorted 83 writers’ answers on how much AI is too much in writing, none of the AI users said they let it write their prose. Both camps drew the line in the same place: fixing a writer’s grammar is a different thing from writing the book. Calling a grammar pass theft erases that line entirely, and it tells every writer who uses a spell checker that they’re a fraud too. Is that really the line you want to hold?
The pro-AI side had its own weak answers. “You’re just jealous” isn’t an argument, and the author it was aimed at had just described her rent. “It’s more work to write with AI than to write it yourself” drew the most ridicule in the whole thread, and deserved it, since the flood in the romance categories exists because it’s so much faster. And the comparison to digital photography, where the old guard scoffed and then everybody switched, skips the part the anti side cares about. Nobody trained a digital camera on pirated photographs.
One comment, from somebody in neither camp, was sharper than anything else in the thread. She said the pro side thinks it can argue people into approving of AI, the anti side thinks it can shame people out of using it, both sides antagonize each other, and the people in the middle get drowned out.
Who gets hurt when every book is accused of AI?
The writers who don’t use it.
Halfway down the thread, one writer said he’d stopped writing. He doesn’t use AI. Every time he posted a book in the group, people said it was AI anyway, and now he’s self-conscious telling anybody he wrote a book at all, because he assumes they’ll think a machine did it. He said AI had destroyed the authenticity of writing, and so had the people in that group.
That one made me angry. The mob in these threads says it’s protecting human writers. I hate that. Here was a human writer telling them they’d driven him out. How many more quit without saying a word?
Another writer said her book was written with no generative AI at all, and she ran it through an AI checker anyway. It came back 19 percent AI.
She asked how. Detectors guess, and they get human writing wrong. How is a writer supposed to defend herself against a guess? If you’ve ever run your own clean chapter through a detector and gotten a number back, you know how sick that feels. I’ve written about why AI detection can’t be treated as evidence, and this thread shows the cost. A community that treats suspicion as proof ends up punishing the innocent at the same rate as the guilty, and the innocent are the ones who take it personally and quit.
I quit writing groups a long time ago, because they were hostile long before AI gave them something new to fight about. Threads like this one are the reason. A beginner can’t ask a question in one without somebody treating it as a confession. If you’ve posted in one, you’ve probably felt it.
So why do people care?
Go back to the original question. If Bob publishes a book written by AI, why does Sally care? What does she share with Bob that he can take from her?
Sally cares if she publishes in the same category and gets paid from the same pool, and if her books were in the pirated collection that trained the model Bob used. As a reader, she cares if Bob’s cover says nothing about how the book was made and she paid for it expecting a person. Those are real stakes, and the thread proved every one of them.
She has no stake in whether Bob used a grammar checker on a book he wrote himself. She doesn’t get to call him a thief for it, and she certainly doesn’t get to call the next writer a fraud because a detector spit out a number.
The poster who asked the question kept replying to people with “thanks for responding.” She asked in good faith and got called disingenuous, stirring the pot, and fishing for permission. She wanted an answer and got a brawl, and the same thing happened to a mother who asked how to publish her daughter’s bedtime story in a group just like it.
Where I stand
I use AI every day as an assistant: transcripts, summaries, research, first-pass outlines, cleanup. I still write the book. I’ve laid out the labor split that works on a book in detail. A book generated by a machine and published under a person’s name as if they wrote it is a lie to the reader, and I won’t do it. Why would I want my name on a book I didn’t write?
The flood is real, and it’s hurting authors who make their living in genre fiction on Amazon. The piracy was real, and a court and a billion-and-a-half-dollar settlement say so. The missing labels are a real failure, and Amazon could fix it tomorrow with information it already collects. On those three, the angry side of that thread is right, and I’d tell anybody asking the original question to take those answers seriously.
So is the mob, and it’s hurting the wrong people. A writer quit. A clean book got flagged by a machine that guessed. A man who plots his own action scenes got told he’s not an author because software fixes his commas. None of that protects a single human writer from the flood.
The vendor who grew her own tomatoes
Back at the farmers market, the fix for the factory tomatoes is a market manager who checks where the produce comes from, a sign rule that means something, and a limit on how many trucks one vendor can bring. Shouting at every vendor with a nice-looking tomato fixes nothing. Amazon is the market manager in this story, and so far it’s mostly collecting fees.
Until it does better, the vendors have to police themselves without turning on each other. Tell buyers what you sell. Call out the lies you can prove. Leave the woman with thirty years of tomatoes alone. Who wants to shop at a market where every vendor is a suspect?
That’s the pattern I keep writing about in the Dogpile series, and this thread is the clearest example I’ve seen of how a fair question turns into a pile-on. If you’re working out where AI belongs in your own writing, start at the AI & Writing Hub, and if you want help drawing that line for a book of your own, my AI services are built for it.
