The seven keyword slots and three category choices in Amazon KDP are among the most consequential positioning decisions a fiction author makes at launch. Most authors spend fifteen minutes on them, pick what seems closest to accurate, and move on. That’s understandable, and it’s a mistake, because these choices decide where Amazon places the book, which genre readers see it in browse results, and what the recommendation system pairs it with.
Across genres, from thriller to romantasy to cozy fantasy to science fiction, your Amazon categories and keywords are not admin. They’re the first and most persistent act of positioning your book makes on Amazon.
What do Amazon categories actually do for fiction authors?
Categories feed the placement signal that Amazon’s discovery system uses to decide who sees your book. A large share of Amazon fiction sales don’t come from readers searching a specific title. They come from recommendation surfaces: “customers also bought,” “you might also like,” category browse pages, and algorithmic email recommendations. Your categories feed the initial signal for those systems, so if the signal is wrong, the system puts your book in front of readers who won’t buy it, or stops recommending it at all.
Amazon also learns from what readers do after a recommendation. If the readers who find your book through a given keyword or category consistently don’t purchase, or buy and then leave ratings that don’t match the expected experience, the system pulls back on that placement. A misaligned choice doesn’t just limit visibility at launch. It erodes visibility over time.
Should you pick broad or niche categories?
Pick the most specific subcategory where your book is genuinely one of the stronger options and where the readers browsing it are likely to enjoy what they find. The instinct is to choose the broadest category that technically fits: write romance, pick Romance. The problem is that broad categories are the most competitive real estate in the store. Getting organic visibility there needs a significant launch spike or sustained ad spend, and even authors who have that will find broad placement doesn’t hold between campaigns.
A specific subcategory with a lower bestseller threshold is more valuable than a broad category you can’t realistically reach the top of. A bestseller badge in a narrow subcategory shows up in search and browse pages and becomes a conversion signal for readers deciding whether to look further.
For thrillers, the distinction carries real weight. A domestic thriller and a fast-paced international conspiracy thriller both fit under “Thriller,” but they serve readers with different expectations and browsing habits. Correct subcategory placement puts the book in front of readers who want what it delivers, and that match is what drives consistent conversion. The question to ask is not “what’s the broadest category that includes this book?” It’s “what’s the most specific subcategory where this book belongs, where the readers there will buy it, and where the bestseller threshold is realistically reachable on my projected launch numbers?”
How should fiction authors choose keywords?
Use your readers’ language, not your own. The way a reader searches for a book is different from how an author describes their own work. An author might call their book a slow-burn enemies-to-lovers fantasy with found-family themes. A reader searches “enemies to lovers dark fantasy,” or “found family romantasy,” or “fated mates slow burn.” Those phrases have different search behavior, and reader language usually drives more traffic. The seven slots should reflect how readers describe the experience of the book, not how the author describes the craft of it.
The most useful keyword research for fiction is qualitative:
- Read the twenty most recent reviews of your book, or of comparable titles if you’re pre-launch.
- Note the exact words readers use for the reading experience: the tropes they name, the emotional language, the comparisons they draw.
- Cross-check those phrases against Amazon’s search autocomplete by typing partial phrases and watching what surfaces. Autocomplete reflects real search behavior and often shows terms you wouldn’t have thought to include.
For books that sit across genre lines, there’s a second keyword set available. A thriller with a strong romantic subplot can use trope-based romance keywords in two or three slots. A science fiction novel with strong relationship arcs can reach both the sci-fi and fantasy-romance communities through different keyword groups. Most authors in genre-adjacent territory treat all seven slots as one genre’s vocabulary and leave the cross-genre audience unreached. The backlist implications are worth thinking through; how backlist marketing works for fiction authors covers the longer-term visibility picture, and metadata is part of what sustains it.
Why does category and keyword mismatch hurt discoverability?
What damages discoverability over time isn’t a slightly imperfect keyword or a partly-right subcategory. It’s a mismatch between what the metadata signals and what the book delivers. Amazon learns from reader behavior after the recommendation. If readers who found the book through a given keyword consistently don’t buy, or buy and then rate it in a way that doesn’t match what that community expects, the system adjusts. The metadata promise has to match the reading experience. Position a book as dark romance when it’s lighter than that community expects, and the algorithm pulls back on the placement because the conversion and rating data tell it the match is wrong.
This is also why copying a comparable title’s categories and keywords without reading what that title actually delivers can backfire. Its metadata works because it aligns with the book and has review data behind it. Borrowing the placement without the alignment doesn’t produce the same result.
A one-hour metadata audit before your next launch
The audit doesn’t take long. Allocate an hour and work through it deliberately:
- Search Amazon for your genre and map the subcategory breakdown. Find where comparable titles are actually placed, not where you assume they belong.
- Check which subcategory bestseller lists are reachable on your projected launch numbers. A subcategory where the current bestseller sits around 2,000 overall is achievable with a moderate launch; one where the top book sits at rank 50 is not.
- Read your best twenty reviews and write down the language readers actually used, then cross-check it against Amazon search autocomplete.
- Test coherence: do your categories and keywords describe the same book? Would a reader who found it through your keywords be likely to enjoy what they got?
If you work with a marketing service on your launch, the metadata conversation should happen in the brief alongside cover, blurb, and pricing, not the day before upload. My launch planning work includes a metadata review as part of the positioning pass, and it’s one of the elements most often skipped when authors manage launches on their own. The results usually reflect that gap directly.
Frequently asked questions
How many keywords and categories does Amazon KDP give fiction authors?
Seven keyword slots and three category choices at upload. Together they determine placement, browse visibility, and what the recommendation system pairs your book with.
How should I choose fiction keywords?
Use reader language, not author language. Read the twenty most recent reviews of your book or comparable titles, note the tropes and phrases readers actually use, then check them against Amazon's search autocomplete.
Should I pick broad or niche Amazon categories?
Pick the most specific subcategory where your book genuinely belongs and the bestseller threshold is realistically reachable. Broad categories are the most competitive space in the store and rarely hold between campaigns.
Can I copy a comparable title's categories and keywords?
Not safely. Their metadata works because it aligns with their book and has review data behind it. Borrowing the placement without the alignment creates a mismatch that Amazon reads in the conversion data and pulls back on.