There are two games in AI writing right now, and most people don't know which one they're playing.

Game one produces a book-shaped object. Something with a cover, a title, an ISBN, and enough words between the covers to qualify. The customer is a consultant who needs a lead magnet, a coach who needs a credibility prop, a person who needs "author" on their LinkedIn bio. The book exists to be had, not read. $97, one click, done. That's a real product for a real customer, and I'm not here to pretend it isn't.

Game two produces a book. Something a stranger picks up, finishes, and thinks about after they put it down. Something where the characters sound like themselves in chapter 22 the same way they did in chapter 3. Something that requires a canon database, a quality scoring system, and an author who won't ship until it passes.

The tools are different. The infrastructure is different. The quality ceiling is different. The entire premise of what "writing" means is different.

The problem is that game one markets itself as game two. "Full manuscript in one click" implies the manuscript is the hard part. It isn't. The manuscript is the easy part. Any model can fill pages. The hard part is making someone want to turn them.

I wrote about this today. A platform that sells AI-generated fiction for $97, with anonymized testimonials, no verifiable reviews, and exactly two sentences of sample output. The sample read like a default prompt response from 2023. They chose it as the showcase.

Self-sort accordingly.