01
Fiction: romance runs away with it
Fiction is where the crowding is most visible. Of 10,467 fiction books, romance is the runaway leader — tagged on 3,400 of them, roughly a third of all fiction. Fantasy and erotica follow well behind, and the field thins steadily from there. This is the clearest supply signal in the dataset: if you are writing romance, you are writing alongside a very large crowd of other AI-assisted authors.
The bottom of this list is as useful as the top. Historical fiction, horror, and crime each sit under 650 books — a fraction of romance's supply. Crowding is not destiny, but if you are hunting for a corner of fiction with fewer AI-assisted authors chasing it, the tail of this table is where to look. To pressure-test any of these against real search demand, the free niche finder surfaces adjacent sub-genres that rarely show up in a top-line genre count.
02
Non-fiction: self-help leads, the long tail breathes
Non-fiction concentrates too, but more gently. Of 3,492 non-fiction books, self-help and personal development lead at 901, with religion and spirituality a clear second. After the top few categories the counts fall off fast — and that long, thin tail is where non-fiction authors have the most elbow room.
Cookbooks (43), true crime (71), and personal finance (106) are barely represented on the creation side — striking, given how commercially prominent those shelves are on Amazon. That mismatch is exactly what a market-timing tool is for; the trend spotter tracks which topics are heating up, and the competition analyzer sizes up who you'd be up against before you commit a single chapter.
03
Visual books: small, early, wide open
Visual books are the newest and thinnest slice — 719 in total, with only a fraction carrying a specific category tag. Children's picture books lead the tagged set, followed by manga and comics. The raw supply here is a rounding error next to romance, which is precisely what makes it worth a look for authors who can produce illustrated work.
Most visual books did not carry one of these specific tags at all, which tells its own story: the category is young enough that conventions haven't hardened. AIWriteBook's visual-book support handles the illustrated formats directly, so an author with a picture-book or comic idea isn't wedged into a text-first workflow.
04
What authors are researching — and what the AI flags as hard
Beyond what people made, we can see what they investigated. Real users ran 480 distinct KDP keywords through our free keyword tool, which returned 3,451 individual competition estimates. This is the closest thing in the dataset to a demand signal — the phrases authors actually cared enough to check.
With that established: the average estimated competition score was 65.2 out of 99, spanning the full range from 5 to 99. Nearly half of all scored keywords landed in the model's "Very Hard" band — a lopsided distribution that says authors are, unsurprisingly, drawn to the same obvious high-competition phrases.
| Difficulty band (AI estimate) | Keywords | Share |
|---|---|---|
| Very Hard | 1,591 | 46% |
| Easy | 494 | 14% |
| Medium | 467 | 14% |
| Hard | 452 | 13% |
| Very Easy | 447 | 13% |
The practical read is not "avoid Very Hard keywords" — it is "most people cluster on them, so the estimated-easy long tail is comparatively under-worked." You can run your own phrases through the same free KDP keyword research tool in a couple of minutes, and pair the terms it surfaces with the categories from our KDP keyword research guide to build a listing that isn't fighting for the single most crowded phrase in its genre.
Reading the map before you write
What the supply side says, with every limitation kept in view:
- Romance is the most crowded creator-side niche by a wide margin. That is not a reason to avoid it — romance readers are voracious — but it is a reason to differentiate hard within it.
- The genre tails are where AI-assisted supply thins out. Historical fiction, horror, and crime in fiction; cookbooks, true crime, and personal finance in non-fiction. Fewer authors making books there does not guarantee demand — check that separately.
- Visual books remain wide open. Low supply and unsettled conventions make illustrated formats a genuine frontier for authors who can produce them.
- Treat AI competition scores as a compass, not a map. Nearly half of researched keywords were flagged Very Hard because authors crowd the obvious phrases — the model's estimates are a relative sorting aid, never Amazon truth. When you plan a multi-book run in whichever niche you pick, the series planner keeps the whole line coherent from the start.
Questions about niche saturation
On the creation side, romance is by far the most crowded — tagged on 3,400 of 10,467 fiction books we tracked, about a third of all fiction. Fantasy, erotica, thriller, and mystery follow. In non-fiction, self-help and personal development lead. Important caveat: this measures how many authors are making books in each niche, not how those books sell — we hold no Amazon marketplace data.
No. Every genre and category count here comes from books created on AIWriteBook — the supply side, not Amazon's storefront. We do not observe sales, rankings, or revenue, and nothing on this page is an income or saturation verdict on the Amazon marketplace itself. It is a map of where new manuscripts are being made.
No — and this matters. Those scores are produced by an AI model prompted to estimate difficulty from the genre alone. They are language-model estimates, useful as a relative sorting signal, and they are explicitly not measurements of Amazon's live search results. Treat them as an informed compass, not as marketplace fact.
On the supply side, the genre tails are thinnest: historical fiction, horror, and crime in fiction; cookbooks, true crime, and personal finance in non-fiction; and visual books almost across the board. Fewer AI-assisted authors are making books there. But low supply is not proof of demand — always validate reader interest separately before committing.
Real users ran 480 distinct KDP keywords through our free tool, generating 3,451 individual competition estimates with an average score of 65.2 out of 99. Nearly half landed in the model's Very Hard band, which mostly reflects that authors gravitate toward the same obvious high-competition phrases rather than the estimated-easy long tail.
Keep reading
This report is part of our Amazon KDP publishing guide, which covers keywords, categories, and the listing craft that turns a niche choice into a findable book. We refresh these supply figures as the dataset grows.