Some questions have a plain technical answer. Ask one in the wrong room, and you can end up as the ground two armies fight on.
An author posted in a Facebook group for beginners self-publishing on Amazon. She had a comic book series about two characters, and she wanted to turn it into a KDP paperback series. Her problem was ChatGPT. Every time she asked it to prepare the pages, it changed the pictures. She’d explained what she wanted a thousand different ways, she said, and it still messed up every image. Did anyone know how to fix it?
That’s a technical question. It has a technical answer.
The first reply told her to learn the craft, hire an artist, learn formatting, work with an editor and watch tutorials, because relying on AI is the antithesis of creativity. She thanked him politely and explained she’d learned to format her first book in Canva from tutorials. Not everyone, she said, can afford an illustrator. A designer told her that not affording an illustrator is poor planning. Another man told her that not being able to afford something doesn’t give her the right to steal it.
It went downhill from there.
You’ve read the Dogpile series. Can you guess how it ended? What made this one different is that the pile-on came from both directions, and the anonymous account defending AI was every bit as ugly as the people attacking it. The author who asked the question was barely in the thread by the end.
What happened in the KDP comic book thread
The thread split into three groups.
The largest group told her to stop using ChatGPT and hire an artist.
“If only there was an entire industry of artists who already know how to create the art you want.”
“Turn off ChatGPT and have a go at actual creativity.” “Nobody wants AI slop.” Was any of that useful to her? A man posted his own hand-drawn pages with the caption “Learn how to illustrate ma’am!” Somebody said the drawings looked like awful slop that should never be allowed near children.
Then one commenter announced that she could fix the problem by not using “the unethical CSAM theft machine.” From that point the thread became an hours-long shouting match between him and an anonymous account defending AI. He called the tool the pedophile machine over and over. The anonymous account called him sick, a creep and a pervert over and over. Neither one said anything about comic books, KDP or the author’s problem again.
The anonymous account didn’t stop there. It went down the thread replying to almost every anti-AI comment with insults: idiots, morons, “you’re not the sharpest tool in the shed,” a string of laughing emojis after each one. Twisting one commenter’s name into a dirty joke. Calling people stupid for having an opinion about their own industry. Did that account think it was helping the case for AI? It wasn’t. Every reply handed the other side another example of exactly what they think AI users are like.
The third group was small. They answered the question.
What was the author asking?
One commenter, a woman who makes coloring books in Canva, read the post carefully. She pointed out that the images said they were illustrated by the author, so the author already had the art. What she needed was a way to put existing pages into the right format for a KDP paperback, and ChatGPT was giving her a different result every time she asked. The author confirmed it. That was exactly the problem.
So most of the thread was answering a question nobody had asked. She didn’t ask whether she should use AI. She didn’t ask for a lecture about the ethics of illustration. She asked how to get her pages into a paperback, and she said a tool wasn’t doing what she needed.
Another commenter questioned whether the art was really hers, pointing to a typo in the first panel that looked like a ChatGPT storyboard. Maybe it was. That’s a fair question. There’s an honest answer to it about disclosure, and I’ll get to it. It still wasn’t what she asked, and it didn’t need fifty people to ask it.
She asked how to format her comic for a paperback. The thread answered a question she never asked. – Richard LoweShare on X
Why does ChatGPT change the picture every time you ask it to resize?
Because it isn’t resizing anything. It’s drawing a new picture.
When you hand a finished image to an image generator and ask for a different size, it doesn’t stretch or crop the pixels you gave it. It generates a fresh image based on what it sees and what you wrote. Faces shift. Clothing changes color. Text in speech bubbles turns to gibberish or picks up typos. Ask a thousand ways and you’ll get a thousand slightly different pictures. That’s what the tool does.
Resizing and laying out finished art is a job for a layout or image program. I generate cover art in Leonardo, and the final assembly, the sizing and the print preparation happen in PaintShop Pro. The AI makes the picture. Another program puts it on the page. Mixing those jobs up is the cause of most of the frustration I see from authors using AI for illustrated books.
If you’re generating new images and need the characters to stay the same from panel to panel, two commenters gave good advice. Create a reference image of each character on a plain white background first, and give that reference to the AI every time. And change one panel at a time. Asking an image generator to redo a whole page usually changes things you didn’t want changed.
How do you format a comic book for a KDP paperback?
The commenter who makes coloring books laid it out, and her method works for any picture book or comic you already have the pages for.
Pick your trim size first. KDP lists the sizes it prints, and comics usually go into one of the larger ones.
Set up a custom template at that size in Canva, Affinity, PaintShop Pro or whatever layout program you know. If your art runs to the edge of the page, add bleed, the extra margin KDP trims off, and set it up to KDP’s published bleed measurements before you place a single image. Keep text and anything important away from the edges and the inside gutter, because the trim isn’t perfectly exact and the binding eats part of the inner margin.
Then drop the images into the template in order, one page at a time. Export the interior as a print-quality PDF. Build the cover as a separate file using KDP’s cover calculator, because the spine width depends on your page count and paper type. Order a proof copy before you publish. Color printing exposes problems a screen never shows.
One commenter insisted that no one but an artist could ever handle gutters and bleeds. The process was far too technical, he said. It isn’t. It’s fiddly and worth getting right. Authors format their own picture books every day. My formatting guide covers the general process.
Do you have to disclose AI-generated images on Amazon KDP?
Yes. KDP asks whether your book contains AI-generated content when you publish. That includes images. If you used AI to create the art, you have to say so, and saying “illustrated by the author” when a machine drew it isn’t honest.
Here the critics had a real point, buried under the noise. If the panels came from ChatGPT, the book should say so, and readers can decide whether they want it. There’s also a copyright problem: in the United States, purely AI-generated images can’t be copyrighted. That matters if you’re building a whole series around characters you want to protect. I went through both issues in detail when another author got dogpiled over a bedtime story with AI illustrations.
None of that requires calling anybody a thief. A sentence would have done it. Calling her one anyway is AI Derangement Syndrome, and the account defending AI had its own version of the same disease.
Why do AI questions in writing groups turn into fights?
AI is one of the most divisive subjects I’ve ever seen among writers. Every question that mentions it gets treated as a vote on the whole thing.
I was shocked by how much vitriol it draws. I use AI every day as a digital assistant. I don’t write with it. I also understand why illustrators are angry. Many of them believe their work trained these tools without permission, and they see the tools being used to replace them. They’re entitled to that grievance.
But is a beginner asking how to format her comic the AI industry? Of course not. Shouting “theft machine” at her doesn’t pay a single artist. Neither does an anonymous account calling everybody morons. It’s two groups of people using a stranger’s question as a stage. That’s the Dogpile pattern every time: the person who asked becomes the excuse, and the crowd performs for itself. I don’t join writing groups anymore, partly for this reason. I’ve written about it in why Facebook writing groups suck and in why writers care so much about AI.
The people who called her a thief for asking how to make a paperback behaved badly. So did the account calling everybody who disagreed a moron. Both sides of that thread owed her an apology, and neither one was ever going to give it.
My answer to all of it is the same. It’s a tool. Use it or don’t, and let other people do what they do. Readers will buy the book or they won’t.
What should you do when your question starts a pile-on?
Take the useful answers and leave. Why stay for the rest?
She did the right thing. She stayed polite, thanked the people who answered, confirmed the one reply that understood her, and didn’t argue with anybody. The coloring-book author offered to help her directly, and she said she might take her up on it. Two helpful people out of dozens. That was the whole value of the thread.
Don’t defend yourself in the comments. Every reply feeds it. If the group is hostile, ask the person who helped you in a private message and get out. And if you’re the one reading a thread like that, answer the question that was asked, or scroll past.
The social media hub collects the rest of the Dogpile series, and if you want help taking an illustrated book from finished pages to a published paperback, my publishing services cover the formatting and the upload. She asked how to get her comic into a paperback. Somewhere in a thread of a hundred comments, one woman told her.
