A Beginner's Guide to Keyword Research for Book Blurbs and Amazon Metadata

Keyword research has a reputation problem. It sounds like something done in a spreadsheet by people selling lawn fertiliser, not something a novelist needs to worry about. But you already do a version of it every time you describe your book to a friend and watch their face to see which words landed. "It's a murder mystery, but funny" gets a very different reaction from "it's a cosy crime novel set in a Welsh village." Same book. Different hooks. One of those phrasings is closer to what a stranger will type into a search bar at eleven at night.

That gap between how you describe your book and how readers search for it is what keyword research closes. It affects two things you control: the words in your blurb and the metadata you fill in behind the scenes.

Start with the words readers actually type

Readers rarely search in the language of literary description. They search in feelings, formats and comparisons. They type "thriller with a twist", "slow burn romance enemies to lovers", "cosy mystery village", or the title of whatever they finished last week plus the word "similar".

So begin by collecting phrases rather than inventing them. Three places are full of them:

• Retailer search bars. Type the first few words of a phrase into Amazon, Kobo or Google Books and read the suggestions that drop down. Those are real queries, ordered by how often people use them.
• Reader shelves. Goodreads and StoryGraph users label books with tags such as "found family" or "morally grey heroine". Those labels are search language in the wild.
• Reviews of books like yours. Reviewers describe the reading experience in plain words: "gentle", "creepy in the best way", "couldn't put it down". Note the nouns and adjectives that keep returning.

Keep a running note. Roughly twenty phrases is plenty for one book.

Two homes for every keyword

Backend metadata

This is the invisible part: the keyword fields in your retailer dashboard, plus your categories, subtitle and series name. On KDP you get seven keyword fields, each with a character limit of around fifty. No human reads them, so they can be slightly clunky. Use phrases rather than single words, because "second chance romance" is a sharper signal than "romance" or "second" on their own. Fill every field.

Don't repeat words already in your title, subtitle or series name; the system already knows those. And don't stuff in another author's name or a competitor's book title. Retailers' terms generally discourage it, and it can get a listing taken down.

The blurb itself

Your description is searchable too, and it is the one piece of metadata a person actually reads. That makes it the overlap point. The strongest two or three phrases from your list should appear naturally in the blurb, ideally in the first couple of lines, where browsing readers decide whether to keep going.

The blurb does two jobs at once: it signals relevance to a search engine and it persuades a human being. When those jobs pull against each other, the human wins. A blurb written for a robot reads like a robot wrote it, and nobody buys that.

How to build a keyword list in half an hour

1. Write five plain sentences describing your book. Ignore elegance entirely.
2. Turn each one into a two- or three-word phrase a stranger might type.
3. Test each phrase in a retailer's search bar. If autocomplete suggests something close, adopt the more common wording.
4. Look at the top results for that phrase. If they're books genuinely similar to yours, keep it. If they aren't, the phrase is wrong for you.
5. Sort what survives into three groups: what it is (genre, subgenre), how it feels or who it's for (mood, trope, theme), and where it sits (the territory of comparable titles, described without naming living authors).
6. Copy the best two from each group into your note. Those six are your working keywords.

Categories are a shelf, not a label

Categories decide which browsable shelves your book appears on, and which bestseller lists it is eligible for. A narrow category you can realistically top is worth more than a broad one where you'll never be seen. "Cosy Mystery > Animals" is a genuine shelf with genuine readers. "Mystery, Thriller & Suspense" is a supermarket aisle where the big names own the front of store.

Look at what already sells in a category before you commit to it. If the top twenty titles are all from major publishers and heavily discounted, hunt for a smaller neighbouring shelf where your book still fits honestly. Never place a book in a category it doesn't belong in. It costs you returns, reviews and goodwill.

Writing a blurb that carries keywords without groaning

If you can't say the phrase out loud to a friend without wincing, it belongs in the backend, not the blurb.

Weak: "A cosy mystery, small town mystery, amateur sleuth novel about a village fête."

Better: "The village fête was supposed to be harmless. Then the parish council chairman dropped dead beside the jam stall, and retired librarian Margaret Bream was the only person who noticed the knife missing from the cake table."

The second version says "village", carries the cosy tone, and delivers an amateur sleuth through character, all without announcing any of it. Use a keyword once. Twice at most. Repetition reads as spam to readers and adds nothing to a search engine that has already understood you.

Your pre-publication routine

Thirty minutes, once, before you hit publish:

• Six working keywords written down and checked against a retailer's search bar.
• Every backend keyword field filled with phrases, not single words.
• Categories chosen for fit and reachability rather than prestige.
• The blurb read aloud, with your strongest phrase sitting in the first two lines.

Then note the date and leave it alone. Come back in three months, look at your impressions and conversion figures, and change one thing at a time so you can tell what moved the needle. Keyword work isn't a one-off task. It's a slow, quiet conversation with the way readers describe the books they love.

Photo: Yan Krukau / Pexels