The Numbers Are Wild, and Nobody Quite Knows What They Mean
It’s 2:47 AM and I’m reading through Amazon’s latest flagging report, and honestly, I need to talk to someone about this. In early 2025, according to Bloomberg’s investigation into Amazon’s AI book flood, the platform identified over 700,000 titles suspected of being AI-generated content. That’s not a typo. Seven hundred thousand. And here’s where it gets dizzy: that represents a 200 percent increase from 2023, which itself felt alarming at the time.
But before you spiral, let’s pause on this number. What does “suspected AI-generated” actually mean? The detection methods are imperfect. Some false positives exist. And yet, the trajectory tells us something real is happening. The sheer volume suggests we’re not looking at a fringe phenomenon anymore. This is infrastructure-level disruption. Self-publishing platforms, which democratized book publishing so beautifully, are now buckling under what might be called the great quantity flood of 2024-2025.
The thing that keeps me awake isn’t just the numbers themselves. It’s what they represent about attention and discoverability. When there are 700,000 additional titles competing for shelf space, even digital shelf space that theoretically has infinite room, something fundamental shifts. The signal-to-noise ratio breaks. And readers like us, readers who want to find genuine work from genuine humans, suddenly need better filters than we ever did before.
Traditional Publishers Draw a Line (Finally)
Here’s what surprised me most: the establishment publishing world actually moved fast. Penguin Random House and HarperCollins, the two largest trade publishers in North America, both updated their author contracts in late 2024. The update was explicit and unambiguous: no AI-generated manuscripts without disclosure. For an industry that moves like a freighter ship, this felt like a speedboat.
What I find interesting about this contract shift isn’t just that it happened, but what it signals about perceived liability and risk. Publishers aren’t banning AI entirely, which is pragmatic and maybe even wise. But they’re requiring transparency. They’re saying: “If you’ve trained on this, if a machine wrote the bulk of it, we need to know.” That’s a different conversation than a blanket prohibition. It’s a conversation about honesty.
In January 2025, The Authors Guild released a formal position paper calling for mandatory AI-disclosure labeling on all published works. Within two weeks, over 15,000 authors had signed on. Fifteen thousand. That’s solidarity on a scale you don’t often see in an industry famous for its competitiveness and solitude. Authors across genres, from literary fiction to romance to sci-fi, said the same thing: readers deserve to know what they’re reading.
You can read the full statement at The Authors Guild AI Disclosure Statement 2025. What struck me most about the language is how it centers the reader. Not the publisher’s bottom line, not the author’s convenience. The reader. You and me. The people who open a book hoping for that authentic human consciousness on the other side of the prose.
But Here’s What Readers Actually Think (and It’s More Complicated)
This is where I have to sit with some uncomfortable complexity. A 2025 study by the Codex Group asked readers directly: would you buy a book if it was disclosed as AI-generated? Forty-four percent said no. They wouldn’t purchase it under any circumstances. That’s nearly half the reading population making a clear ethical or aesthetic choice. But the other half? Twenty-eight percent said they were indifferent. They don’t care who or what wrote it, only whether the book works for them. The remaining readers fell somewhere in the murky middle, uncertain or conditional in their responses.
I keep returning to that 28 percent figure. It’s haunting me in a way I didn’t expect. Because I understand them, even as I disagree with them. If a book moves me, if it makes me feel seen or teaches me something true about the human experience, does the origin of its creation matter? There’s a real philosophical question embedded there, not just a commercial one.
Yet I also know this: when I read a sentence that stops me cold, when I notice the exact weight of a word choice or the rhythm of a particular construction, I’m often feeling the presence of a mind working through a problem. Sometimes that problem is narrative. Sometimes it’s existential. Either way, there’s intention there. There’s a self. And I’m not sure I’ve felt that same quality of intentional self-awareness in the AI-generated excerpts I’ve read. The prose is competent. Often it’s smooth. But it reads like someone reciting grammar rather than thinking through language. That distinction matters to me. I can admit it might not matter to everyone.
The Slush Pile Is Collapsing
Here’s the part that actually worries me most, the part I think about when I’m considering what this means for future literature. The Science Fiction Writers of America reported that short fiction submission quality assessments dropped 31 percent in 2024. Thirty-one percent. And slush pile readers, the underpaid heroes who wade through hundreds of submissions monthly, attributed much of this decline to an influx of AI-assisted entries.
Think about what this means practically. Literary journals and small magazines, already operating on threadbare budgets, now have to spend more time filtering noise. That’s time not spent developing relationships with emerging writers. Time not spent reading the genuine experimental work that might change how we think about the form. When the signal-to-noise ratio degrades, the cost of finding quality rises, and only institutions with resources can bear that cost. Suddenly the path to publication gets narrower, not wider.
This is where the AI debate stops being abstract. It becomes a question about gatekeeping and access. It becomes a question about what kinds of voices get heard and what kinds get drowned out. The democratization that self-publishing promised was supposed to make room for more people. Instead, we might be watching that room get packed with so much stuff that the actual humans trying to enter get stuck in the doorway.
So What Now? What We Read Matters
I don’t have a clean resolution to offer you here. This isn’t a problem with a tidy answer. But I think we have some agency in how this plays out, even if it doesn’t feel like it from inside the chaos.
First: transparency matters. When you’re considering a purchase, when a book is disclosed as AI-generated, you now have information you didn’t have before. You can make your own choice. Whether that’s a dealbreaker or irrelevant or somewhere between is genuinely up to you. But you get to choose with full information.
Second: support the work of writers you love. Review their books. Recommend them to friends. Tell the algorithm, in whatever language it speaks, that you value human-made literature. That attention and care are more valuable than you might realize in a system predicated on volume.
Third: engage with the sentence. Really read. Notice when something feels authentic and when something feels constructed. Pay attention to the subtle work that happens at the micro level, where I think much of what makes literature matter still lives. That close attention is its own form of resistance to meaninglessness.
What’s your instinct on all this? Have you noticed a difference when you’re reading? I’d genuinely like to know what you think, because I don’t believe there’s one right answer here. There’s only the ongoing conversation between writers and readers about what we owe each other when we meet on the page.
