AI Detection Won't Save You From Amazon, and Neither Will Running Away

Seventeen DMs before coffee. Seventeen! About the same Substack post. This better be good.

Brian Niemeier published a piece arguing that AI detection will soon let Amazon auto-label every book as human or machine-written, collapsing indie publishing in the process. His prescription? Abandon Amazon and embrace something called “neopatronage.”

Some of this is decent advice wrapped in a panic attack. (Sooo much panic.)

Detection Still Doesn’t Work

The foundation of Niemeier’s argument is an anonymous source. “One well-placed tech worker in a Fortune 500 company” says the detection tools in beta right now are “uncannily accurate.” Cool. I’ve also heard from well-placed tech workers who said self-driving cars would be everywhere by 2020.

Detection tools have been “about to get really good” for two years now. They keep improving, sure. They also keep flagging human writing as AI-generated. Niemeier himself acknowledges that War and Peace gets flagged as machine-written, then waves it away by saying the next generation will be different. (Sure, Jan.)

I’m not restructuring my career around “trust me bro, the tech is coming.”

And even if detection improves dramatically… there’s a gray area nobody has solved. What about the author who brainstorms with Claude, then writes every word herself? The one who drafts by hand but uses AI to sharpen her dialogue? Where does a binary “Human-written / AI-generated” label land for those people?

Nowhere useful, I imagine.

Amazon Is Amazon

Niemeier’s argument needs two things to be true at the same time, and they cancel each other out.

Thing one: readers hate AI-generated content. When detection labels appear, they’ll see how much of the catalog is AI, trust will crater, and they’ll flee. His exact words: “when readers see just how many authors are really publishing A.I. slop, their trust in newpub is going to crater.”

Thing two: AI publishers will dominate Amazon because “scale always wins.” They’ll flood the store with content, bury human authors, and Amazon won’t care because indie authors “have never been a profit center” anyway.

But… if readers hate AI books, those books don’t sell. And Amazon’s algorithm promotes what sells, not what exists. Ten million AI books nobody buys isn’t scale Amazon wants. It’s a storage cost. So the flood he’s worried about drowns itself.

And if readers don’t actually care (which, tbh, most don’t until someone presents it as a moral issue), then the labels are irrelevant and there’s no trust crater.

He can’t have both. Either readers reject AI and the market self-corrects, or readers don’t care and the labels don’t matter.

There’s also a weird thing where he gets his own conclusion backwards? If detection labels exist and readers genuinely prefer human writing, then labels help human authors. Readers can now filter for exactly what they want. That’s a competitive advantage indie authors have never had before. Niemeier argues labels will destroy indie publishing, but his own premise says the opposite.

”Neopatronage” Is Direct Sales in a Toga

Niemeier presents “neopatronage” as a revolutionary response to AI detection, but… authors have been doing this for years? Mailing lists, direct sales, Patreon, building off-Amazon revenue streams. Indie authors figured this out ages ago. Renaming it doesn’t make it new, and tying it to an AI detection panic doesn’t make it urgent.

If his Amazon argument held up, maybe there’d be a reason to treat this as some kind of exodus moment. But he’s telling authors to flee a platform based on a contradiction. That’s not strategy, it’s a fire drill with no fire.

And the discovery problem gets worse when you leave Amazon, not better. Every author building their own isolated kingdom means readers pick which castles to visit and ignore the rest. Co-ops could help, but co-ops are basically just new Amazons waiting to happen. Aggregation is how we got here.

Diversify because single-platform dependence has always been risky. That’s it. That’s the whole insight, and it predates AI by a decade.

The Assumption That Ruins Everything

Anyway, all the above is nitpicking compared to the real problem with Niemeier’s post.

He assumes AI-assisted books are bad books. His entire framework depends on it.

Detection exposes AI content, readers recoil, the market fractures into “authentic writing” and “slop.” He uses phrases like “assembly line” and “industrial-scale content” as if every manuscript that touched an LLM came out as incoherent garbage.

Some of it does. I’m not going to pretend otherwise. The get-rich-quick crowd has always existed in self-publishing. AI just gave them a faster printing press. Those people suck. They sucked before AI, too, when they were paying $200 for ghostwritten books on Fiverr and keyword-stuffing their way onto bestseller lists. The tool changed. The grift didn’t.

But across four thousand words, Niemeier never once considers that AI might make books better.

The authors I talk to are using AI to brainstorm plot structures they wouldn’t have found on their own. To stress-test worldbuilding for consistency holes. To iterate on dialogue until it sounds like their characters instead of “close enough” at 1am when their brain is mush. To push through revision hell (and oh my god, revision hell) so they can spend more energy on the creative decisions that actually matter.

Those aren’t lazy shortcuts. They’re better tools in the hands of people who already know what good writing looks like.

Niemeier frames every AI-touched book as a product of laziness, and the opposite is just as common. A writer who cares deeply about her craft now has a tool that lets her care more effectively. Super different picture than the one he paints.

Better Books Win

Reading rates have been declining for years. The easy explanation is phones and short attention spans, and yeah, that’s part of it.

But maybe publishing also hasn’t given people enough reasons to pick up a book? Writing has always been brutally expensive in time and energy. Talented authors compromise because they hit a wall of exhaustion before the deadline. Promising ideas die in draft two because the author couldn’t crack a structural problem alone, and nobody was around at midnight to help think it through.

AI changes that equation. A skilled author working with AI can iterate more, catch problems earlier, and pour energy into the weird creative leaps that make a story worth losing sleep over instead of drowning in mechanical friction. (Mochi, my fluffy-butt cat, handles a different kind of friction by parking herself on my keyboard at 2am, but we manage.) ;)

I genuinely think we’re heading into an era of better fiction. More ambitious plots. Richer characters. Ideas that would’ve died in a drawer a few years ago actually reaching readers. The gap between a good book and a great book has always been revision, and AI compresses the cost of revision dramatically.

Harry Potter didn’t just sell a billion copies. It created a whole generation of readers who barely touched books before. Great stories do that. If AI helps skilled authors write fiction that keeps people up past their bedtime, everybody wins. And I’m all here for it.

Gimme awesome books!