A few days ago someone in a Facebook writing group posted a question and called it rage bait before anybody else could. He asked how much AI is too much in writing. He asked everyone to keep it civil and admitted that was asking a lot. He dared people to get through the thread without the word “slop.” Then he confessed he’d run his own post through ChatGPT. He struck the match, then handed out fire extinguishers.
I don’t join writing groups anymore. They turn into troll city, and whoever posts ends up the target. This group is worse than most. It’s packed with anti-AI trolls, and a writer who admits to using any AI there gets piled on. But this thread landed in front of me, and I read it top to bottom. Then I counted. What I found made me laugh, and then it made me mad. The writers in that thread agree with each other. Nearly all of them drew the same line in the same place, and they spent the afternoon calling each other thieves and cavemen anyway.
The potluck test
Picture a church potluck. Somebody sets a cake on the table. Nobody asks whether she used measuring cups. Nobody cares if she found the recipe in a cookbook or phoned her sister to ask what to bring. A purist or two might raise an eyebrow at the stand mixer.
Everybody at that table agrees on one thing, though. If she bought the cake at the grocery store, peeled off the sticker and slid it onto her own plate, she didn’t bake it.
That’s the whole AI argument in that thread. Almost nobody defended the store-bought cake. They fought about the mixer, and they fought as if the mixer and the grocery store were the same crime.
How much AI is too much in writing, according to 83 writers?
I sorted every top-level comment. I left the replies out, because most of the replies were the same dozen people circling each other. That left 83 answers.
The biggest group, 30 people or 36%, use AI for something past spell check. They use it for research, for brainstorming, for sorting notes, for a second opinion on a draft. Most of them said, a few in capital letters, that it doesn’t write their words. Another 19 (23%) use no AI at all, and several of them said any amount is too much. Thirteen (16%) run a spelling and grammar checker and stop there. The rest cracked jokes, posted memes, stated a rule without saying what they do, or went after the question itself.
Take out the jokes, the links and the people who only came to scold, and 62 writers told us what they do. Forty-three of them use some kind of AI. That’s 69%, close to seven in ten.
Remember where those numbers came from. This group leans hard against AI, and its regulars treat any admission as a confession. A writer who uses AI and wants to avoid a beating stays quiet in a thread like this one, and I don’t blame anybody who did. So the no-AI share is inflated and the AI share is a floor. In a room that didn’t punish the answer, I’d expect far more hands to go up. Even here, the number lines up with the surveys I’ve written about, where 61% of writers use AI in some part of their work.
One number keeps pulling me back. Of those 43 writers, not one said they let the machine write their prose. One account in the entire thread promoted AI-generated books. That’s the total: one out of 83, and it was an ad.
Where do writers draw the line on AI?
I went back through the 43 users and wrote down every task they named. People named more than one, so the percentages overlap. The pattern jumped off the page.
Spelling and grammar came first, at 58%. Brainstorming and sorting ideas came in at 44% and research at 42%. Then the floor drops out. Only 14% let AI give feedback on a draft. Two people, 5%, let it suggest rewording for their own sentences. Writing the prose itself came in at zero.
About half of the users went out of their way to say AI never writes their sentences. A hard science fiction writer put it in capital letters: don’t let it write, because it’s terrible at writing. One man tells Gemini to point out his spelling and punctuation mistakes and orders it never to rewrite anything, so none of someone else’s prose sneaks into his. Another writer asked a chatbot for scene ideas and got suggestions so far off base they sent the writer straight back to their own. These days the chatbot holds the notes. That’s all it does.
That’s the cliff. These writers trust the machine to wash the dishes and fetch the groceries. They don’t let it near the stove.
The fight was over one word
The thread kept tripping over two letters. To one person “AI” means ChatGPT cranking out a novel. To the next it means the red squiggle under a typo. Those are different things, and the thread treated them as one.
A few people tried to untangle it. One writer rated her own use as zero generative AI and assistive AI limited to the spell checker, and she got told grammar checkers aren’t accurate. Somebody else said that anyone who has used a word processor has used AI. A reply shot back that assistive AI and generative AI aren’t the same and that claiming everyone uses AI is false. Both of them were right. They were talking about the mixer and the grocery store, and neither one noticed.
Put the two camps side by side and look at the middle.
The no-AI writers said the prose has to be human. So did the AI users. The no-AI writers said AI writes bad prose. The science fiction writer with dyslexia said the same thing louder. Several people on both sides said a human editor catches tone and subtlety that no program will. The overlap in that diagram covers the question the thread asked. Everything outside the overlap is about something else: stolen training data, the price of an editor, how a person learns to write.
And still the thread burned. That’s what gets under my skin. Two groups of writers agree on the thing that matters most, and they couldn’t hear it because “AI” set off the alarm before anybody finished reading.
Why do writing groups attack writers over AI?
Because the word works like a flag, and once somebody plants it, everybody picks a side before they read the post. In a group where the anti-AI trolls run the comment section, the flag goes up fast and the pile-on starts before anyone checks what the writer said. The worst case in the thread didn’t happen in the thread. A writer showed up to describe what she’d been through in that same group over the past few months.
She posted that she’d finished her book. Somebody replied that she must have used AI to finish it, because it took her so long. Follow that logic. If she’d used AI, she’d have finished faster. She pointed that out herself, and I can’t improve on it.
Later she posted that she was turning the book into a TV show. One person called her antisemitic over a word she never used. Another told her she isn’t a writer because she uses Grammarly. Her answer was that fine, she’s a storyteller, then. She ended with a plea anybody who’s been piled on will recognize: talk about what I posted.
That’s a dogpile, and it follows the same script as the mother who asked how to publish her bedtime story. Somebody asks a plain question or shares plain good news. The crowd hears “AI” and stops seeing the person.
It happened in the thread too, in smaller doses. A woman said she’d never accept AI in any form. A man told her to reword her comment. That didn’t help. When he explained that he uses Gemini to flag spelling errors and never lets it rewrite a word, she told him to stop talking to her. Someone posted that no AI is the best AI and got a caveman impression in reply. Another commenter compared the question to asking how much stabbing is too much stabbing. A newcomer to the group got told that since he’d admitted running his post through ChatGPT, he must use it for everything. Nobody who uses it, she added, is a writer. Both sides threw punches. The pro-AI side called the other side cultists and snobs, and the anti-AI side called the other side thieves.
None of it moved anyone an inch. Would you change how you write because a stranger drew you as a caveman? Would anybody?
The reasons that got buried
Eight of the 43 users, about one in five, gave a reason that had nothing to do with laziness. Four have dyslexia. Two write in English as a second language. Three talked about what an editor costs, and one of those three also has dyslexia.
The science fiction writer has dyslexia and dysgraphia and has been writing since the 1980s. He uses AI as a first editor and says it saves him about $5,000 a year in editing fees. Without it, he’d pay his human editor twice as much and write fewer books. He still pays her. He still writes every word.
A writer with a master’s degree in creative writing said they’d paid $3,500 for an editor once and wouldn’t do it again. They use ProWritingAid to catch spelling errors and repeated sentence openers, and the ideas and the writing are theirs. A writer who pays human editors said she can’t easily afford them either. She saves up for a long time, and she made a point of saying she wasn’t judging anyone who uses software instead.
A novelist in Finland writes in Finnish, then translates his own work into English. He runs proofreading software to catch the grammar and idioms he misses, and he checks that the English doesn’t drift from the Finnish. He also ran an experiment I love. He gave an AI the same premise and characters he’d picked for a science fiction novel, then wrote his own version without reading the machine’s until his was done. In places the AI wrote more colorfully than he did, and it showed him his weak spots. He also saw where it fell apart, and he came away certain he’ll never let it write any part of a novel for him.
Then there’s the historian. She has 25 years of period newspapers, magazines and letters, all 160 to 200 years old, thousands of pages. She loads them into a notebook tool that can only answer from her own documents, and she asks it where she saw a fact she half remembers. That isn’t a machine writing her book. That’s a filing clerk who never gets tired.
And my favorite: a writer set up his smart speaker to insult him while he revises. He asks why he does this to himself, and it tells him it’s because he’s a masochist with delusions of grandeur, sir. He laughs, the misery breaks, and he keeps working. I hope he never turns it off.
None of these people are cheating. They’re using a tool to get their own words onto the page. When somebody calls them thieves, the insult lands hardest on the writers who can least afford a human helper.
What the no-AI writers get right
The anti-AI side isn’t a mob of cranks. A traditionally published author in the thread said twenty of her books were taken to train the language models. That happened to thousands of authors, and a federal judge approved a $1.5 billion settlement against Anthropic over pirated books used for training. Her anger is earned. The contracts she signs state that she hasn’t used AI, and she keeps that promise.
Another writer laid out the water and power that data centers use and named a long-running fight near home in Oregon. Those fights are real. A third put the craft argument in two sentences: writing is a learned skill, and a tool that removes the need to learn it removes the skill. I’ve made a version of that argument myself, in the piece on what ChatGPT does to your brain. If you let a machine think for you, you get worse at thinking.
The no-AI writers are right that stolen books are stolen. They’re right that practice is how you get good. Where they go wrong is aiming all of it at the writer with dyslexia who runs a spell checker. She didn’t train the model. She didn’t build the data center. She’s trying to get her story out of her head, same as them.
Where I draw the line on AI in my own writing
I use AI as a digital assistant in my writing. It transcribes my interviews, summarizes what I’ve collected and gives me a first-pass outline to argue with. The writing is mine. I’ve authored more than 113 books and ghostwritten more than 54, and every sentence in them came out of a human head.
So I draw the line in the same place those 83 strangers did: at the prose. I’ve written about why every writer should treat AI as a digital assistant, and I stand by it. I’d add two things the thread skipped.
First, if the machine wrote it, say so. KDP asks, readers deserve to know, and I’ve laid out what you owe them. Grammar help doesn’t need a disclosure. A machine-written chapter does.
Second, quit accusing people. “It took you so long, so you must have used AI” is detection by gut feeling. Even the detector software can’t serve as evidence. You don’t know how somebody wrote their book. You weren’t there. Unless you’ve got proof, keep the accusation to yourself.
Bring the cake you baked
The man who started the thread asked how much AI is too much. Eighty-three writers answered in a group stacked against AI, and the ones who described their habits still put the line at the same spot: the words on the page. They disagreed about spell checkers, research and brainstorming. They didn’t disagree about the prose. If that group had spent half the energy on that agreement that they spent on insults, it would’ve been one of the most useful threads I’ve read all year.
If you’re sorting out where AI belongs in your own work, the AI & Writing Hub lays out what these tools do well and where they break. If you want a human in your corner while you build the skill, writing coaching is what I do.
And the next time you’re at the potluck, don’t interrogate the woman about her measuring cups. Don’t accuse her of buying the cake because it took her a year to bake it. Bring the cake you baked, set it on the table, and let her do the same.
