The debate about AI in literature seems to oscillate between two extremes: AI will destroy literature, or it's just spellcheck with better PR. I've been there, too. But neither is correct, and here’s why:
The book market does not exist in isolation. Other creative industries have already walked this road: music, stock photography, visual arts. They all show the same pattern.
The Five Phases (and Where Books Are Right Now)
Every creative industry that's undergone AI disruption follows the exact same sequence.
Phase 1: Novelty. "Look what AI can do!" The outputs are impressive but obviously flawed. Incumbents dismiss it. The public debates whether AI can "really" create art.
Phase 2: Flood. Production costs collapse. Volume explodes. Quality is uneven but improving fast. Platforms struggle with curation. Discovery gets harder for everyone. AI-related revenue at Shutterstock surged to $104 million in 2023, but failed to offset stagnant traditional licensing income. Platforms made money. Photographers didn't.
Phase 3: Indistinguishability. This is the critical threshold. 97% of music listeners now report being unable to distinguish between human-composed and AI-composed music. Let that sink in. Ninety-seven percent. Once you can't tell, you can't charge a premium for being human-made. The whole "artisanal, hand-crafted" defence collapses.
Phase 4: Platform Value Capture. Total market revenue grows. Individual creator income shrinks. Global recorded music revenues climbed to $29.6 billion in 2024. Music creators will see 24% of their revenues at risk by 2028. The industry gets bigger. Amazon, Spotify, and Netflix capture the surplus. Individual musicians get poorer.
Phase 5: Bifurcation (but not the one you expected). The market doesn't split into "AI vs. human." It splits into "relationships" vs. "commodity." In music, direct-to-fan revenue grew 30% in 2023, while streaming royalties per play kept falling. The winners aren't the "human-made" artists. It’s the ones with direct audience relationships who monetise through multiple channels, regardless of how the music was made.
Yes, you read that right: Regardless of how the music was made.
Books are somewhere between Phase 2 and Phase 3 right now. Nearly half of the 1,279 authors surveyed in mid-2025 reported using generative AI for writing, more are considering it.
What This Means for Readers
You are almost certainly reading AI-assisted books without knowing it. Some of them may be better than they would have been otherwise. A writer with a brilliant premise but shaky prose can now produce something polished enough to deliver on the promise of their idea.
But here's what matters more: discovery has already gotten harder, and it's about to get much worse. Amazon's recommendation algorithm favours frequency and velocity. It surfaces authors who publish fast, not authors who write the most psychologically complex, beautifully crafted books no one has ever heard of. In a market where AI-assisted authors can publish a book a month, a week, or even every day, the writer who puts out one deeply considered novel a year becomes algorithmically invisible. Your favourite slow-burn author may simply vanish from your feed. The algorithm drowned them.
If you care about finding books with genuine depth and a distinctive voice, you need to become a more intentional reader. Follow authors directly. Subscribe to their newsletters. Join their communities. Don’t outsource your discovery to an algorithm.
What This Means for Writers
If you're a writer, the music industry data should be tattooed on your forearm.
The Kindle Unlimited parallel to Spotify is almost exact. Streaming made the music industry bigger than it's been since 1999, but all that profit hasn’t spilled over to musicians. KU is heading the same direction: per-page-read rates will decline as the content pool expands, which means authors who depend on KU royalties will see their income shredded even if their readership stays stable. Same readers. Same pages read. Smaller slice of the payout pool.
Authors whose income is 85% from book purchases through retailers are the book equivalent of musicians who depend entirely on Spotify streams. It works today. It's structural suicide by 2028. The data is unambiguous: platform-mediated income is compressing while direct-to-fan income is growing (30% in 2023 alone). Every percentage point you shift toward owned channels reduces your exposure to platform compression. In case you were wondering why I keep harping on about our Unravelogue community: now you know.
The "creative direction" skill becomes the scarce asset. A photography industry analysis put it bluntly: professionals will need a strong authorial voice so that their style is distinctive. Generic style without volume is no longer commercially viable. That was written about photographers, but it's the single most important prediction for authors, too. What makes good AI-collaborative work good isn't the AI, it's the human's ability to direct it. Their taste, psychological insight, and sense of what's emotionally true. That skill is rare, and it's getting rarer in relative terms as AI makes production trivial. You can teach AI to write a sentence. You can't teach it to understand the mess of being human. Yet.
Volume matters, but not the way you think. Two books a year versus twelve isn't transformative for revenue per title. But it's significant for algorithmic visibility. A new release creates a brief window of increased visibility. Two books a year gives you two windows. In a market where organic discovery is degrading, those windows are your only reliable path to new readers through Amazon. But twelve books of mediocrity won't save you either because that's the strategy AI content farms are already running at scale. You either out-volume them (and formula fiction can genuinely work as a business model) or you go deep on emotional complexity and build a direct author-reader relationship. Ideally both, if you’re superhuman.
The Split That’s Coming
The most common prediction is that the market will split into "AI books" and "human books." It won't. The actual split, the one every disrupted creative industry eventually lands on, is between commodity content and relationship-based content.
Do not try this split at home, kids
Commodity content is pretty much interchangeable. It serves as entertainment, information, or escape, and the specific creator is irrelevant. If you read a book because the algorithm served it to you, enjoyed it well enough, and couldn't name the author three days later, that was commodity content. There will be a vast, growing, increasingly AI-generated ocean of it. It will be competent. It will be cheap. It will make a lot of money for platforms and very little for individual creators if they don’t crank out books at volume.
Relationship-based content is the opposite. You read it because you care about this particular writer's perspective on the world. You follow their newsletter. You pre-order their books. You'd pay more for their work, specifically. You're not just a content consumer, you' are in a relationship with a creative mind.
The visual arts have already tried the "AI-Free" certification route. "Books by People" certification exists. None of it has produced a commercially viable separation at scale, because the underlying problem persists: if the audience can't reliably tell the difference, the certification becomes a tribal marker rather than a quality signal.
The writers who maintain or grow their income over the next five years will be those who either produce at volume with AI or those who build relationships with their fans. Not by waving a "human-made" flag, but by being so distinctively themselves that their audience isn't just looking for "a good book," they're looking for your book.
The Uncomfortable Bottom Line
If you're a reader: The books you love aren't going away, but the systems that help you find them might not work any longer. Take ownership of your reading life. Build direct relationships with the writers whose work matters to you.
If you're a writer: the book market will not collapse. But your income within it probably will, unless you diversify away from platform dependence and lean hard into the thing AI can't do. Have a genuine perspective on what it means to be human, with all the mess and contradiction that entails. Or go for high volume and AI generation. Either path is valid.
Until next time,
