Last reviewed by Robert Prime — March 2026
UK note: UK-specific considerations apply — ISBN purchases go through Nielsen (not Bowker), VAT rules differ from the US (print books are zero-rated; ebooks carry 20% VAT), and GDPR applies to any email/customer data. See our UK self-publishing guides for specifics.
KDP lets you enter 7 backend keywords for each book — phrases that help Amazon's algorithm match your book to readers searching for it. They're invisible to customers but central to discoverability.
Plenty of new writers fill these in once at upload and never touch them again. That's a mistake. Good keyword research can double a book's organic discoverability; bad keywords (or single broad terms) can leave the book invisible.
This guide covers how the system works in 2026, where to find the right phrases, and how to iterate.
What KDP keywords actually do
When you upload a book, KDP's "Keywords" section gives you 7 fields. Each field holds one keyword phrase (up to ~50 characters).
Amazon's search algorithm uses these phrases (alongside title, subtitle, category, A+ content) to:
- Show your book in organic search results when a reader searches for those phrases.
- Match your book to "Customers who searched for X also bought" carousels.
- Determine eligible categories you can request beyond your two listed ones. Specific keywords unlock niche sub-categories.
- Inform paid-ad targeting eligibility. Sponsored Products ads use these keywords as one signal for ad placements.
The keywords aren't visible to readers but they shape what readers see.
Phrases beat single words
The single biggest mistake: filling the 7 slots with single broad words.
Bad:
- thriller
- mystery
- crime
- detective
- suspense
- police
- murder
Good:
- british police procedural thriller
- cold case detective mystery
- small town crime series book 1
- female detective murder mystery
- yorkshire detective novel
- slow burn psychological thriller
- atmospheric british mystery
The good version targets real searches. The bad version targets head terms where you'll never compete with traditionally-published bestsellers.
Why phrases win:
- Real users type phrases. Nobody searches "thriller" expecting useful results.
- Less competition. Tens of thousands of books target "thriller". Far fewer target "british police procedural thriller".
- Higher intent. Someone searching for "small town romance British" is a buyer, not a browser.
How to find your 7 keywords
Step 1: brainstorm 30-50 candidate phrases.
For a small-town British detective mystery, candidates might be:
- british detective mystery
- police procedural british
- yorkshire detective novel
- small town british mystery
- cold case british mystery
- female detective mystery
- slow burn detective novel
- contemporary british crime
- british cozy mystery
- village crime mystery
- amateur detective british
- gritty british crime thriller
- british murder mystery series book 1
- atmospheric british mystery
- ...etc
Don't filter yet. Just brainstorm. Aim for 30+.
Step 2: validate each with Amazon's auto-suggest.
Open amazon.co.uk in an incognito browser. Type each candidate phrase one character at a time. Amazon auto-suggests completions — those are real searches readers do.
If your phrase auto-completes (or is close to one), readers are searching for it. If nothing auto-completes after typing the whole phrase, the search volume is low.
Step 3: check competition.
For each surviving candidate, search the phrase and count how many results Amazon shows. Anything under 5,000 results = realistic competition. Over 100,000 = head term territory.
Step 4: tools (optional but useful).
- Publisher Rocket (£99 one-off) — searches Amazon's database and gives competition scores + estimated monthly searches per phrase. The single most-used indie keyword tool.
- KDSpy (£35-£60/month) — focuses on category research but useful for keywords too.
- Ahrefs / SEMrush — general SEO tools occasionally useful for finding adjacent search terms, but expensive.
Step 5: pick your 7.
Selection criteria:
- Mix of head terms (broader, more volume, more competition) and long-tail (narrower, less volume, less competition).
- All 7 must be relevant to your actual book. Stuffing irrelevant keywords gets you penalised.
- No duplicates. Don't use both "british mystery" and "british mystery novel" — pick one.
The 7 keywords structure that works
For a typical genre novel:
- 2 head terms (high competition, high volume) — get you into broad searches
- 3 mid-tail phrases (moderate competition, moderate volume) — the workhorses
- 2 long-tail phrases (low competition, niche fit) — easy wins where you can rank quickly
Where keywords show up
Beyond the 7 backend slots, keyword phrases also need to appear in:
- Book title (Amazon weighs title keywords most heavily)
- Subtitle (second-highest weight)
- Book description (third-highest)
- A+ Content (lower weight but still matters)
- Editorial reviews / Author Central (low weight)
A book whose title is "Cold Case" but whose 7 keywords are all about "psychological thriller" will struggle. Title + keywords need to reinforce each other.
Avoiding keyword stuffing
Amazon's terms prohibit:
- Keywords irrelevant to the book ("free", "bestseller", "Stephen King")
- Other authors' names as keywords (especially current bestsellers — Amazon will reject the keyword)
- Misspellings as a hack ("hary potter")
- Time-sensitive terms ("2026 new release") — these decay fast
- Trademarked terms ("Kindle", "Amazon")
- Subjective claims ("best", "amazing")
Books with stuffed keywords get either deindexed or have the keywords stripped silently. Stick to relevant phrases.
Iteration — the bit nobody does
After 90 days of being live, every book should have its keywords re-evaluated. Two data sources:
1. Amazon Ads search-term reports.
If you're running Sponsored Products ads on auto, Amazon will generate a search-term report listing every actual reader search that triggered your ad. The high-converting searches are your real keywords — often different from what you guessed.
Update your 7 backend keywords with the top performers from search-term reports.
2. KDP Brand Analytics (if eligible).
For brand-registered authors, Brand Analytics shows search frequency rank for any phrase. Useful for confirming your keyword candidates have real volume.
A 90-day keyword refresh cycle is standard for serious indie authors. Set a reminder.
UK considerations
- Amazon.co.uk and amazon.com have separate keyword indexes. A keyword optimised for US readers may not auto-suggest in the UK. If you publish in both markets, you can set different keywords per marketplace — but in practice, most authors share keywords and accept some loss.
- UK English spelling matters. "Color" vs "colour", "behavior" vs "behaviour" — UK readers search with UK spelling on amazon.co.uk. US readers do the opposite.
- British setting / character keywords have strong commercial value. "British detective", "yorkshire mystery", "scottish thriller", "cornwall cozy" all index well on amazon.co.uk and increasingly on amazon.com (the British-setting market is large in US too).
Keyword fields by genre
A general framework — your specific phrases will vary:
Romance:
- Sub-genre + tropes (e.g. "small town second chance romance")
- Heat level (e.g. "clean romance" or "spicy romance")
- Setting / era (e.g. "regency romance")
- Series book number (e.g. "billionaire romance book 1")
Thriller:
- Sub-genre (e.g. "psychological thriller" or "domestic thriller")
- Protagonist type (e.g. "female detective thriller")
- Setting (e.g. "british crime thriller")
Fantasy:
- Sub-genre (e.g. "urban fantasy", "epic fantasy")
- Magic system / world (e.g. "necromancy fantasy")
- Protagonist (e.g. "female assassin fantasy")
Non-fiction:
- The exact reader problem ("how to start a podcast")
- Audience descriptor ("for beginners", "for UK authors")
- Outcome ("step by step", "complete guide")
Common mistakes
- Single broad words. "thriller" is not a keyword.
- Never updating. Set-and-forget is leaving money on the table.
- Ignoring Amazon Ads data. Search-term reports tell you your real keywords for free.
- Using competitor author names. Amazon strips these.
- All 7 keywords very similar. Diversify so you cover the breadth of how readers describe your book.
- Keywords that don't match the cover or blurb. Mismatch tanks conversion when readers click through.
The bottom line
Pick 7 phrases (not single words) that real readers type. Use Amazon's auto-suggest to validate. Mix head terms, mid-tail, and long-tail. Re-evaluate every 90 days based on Amazon Ads search-term data.
This is the cheapest, highest-leverage marketing work you'll ever do — and the one most authors never bother with.
See whether AI recommends your book
Readers increasingly ask ChatGPT, Claude, Gemini and Amazon’s Rufus “what should I read next?” — and the AI names a shortlist. Our AI Bookshelf Report found AI cites Wikipedia 16× more often than Amazon when recommending books, so the listing levers in this guide are only half the picture.
Check your book’s AI visibility free → — see which engines name your book, in ~90 seconds.
Related guides
Frequently asked questions
What's the most common mistake first-time authors make with kdp keyword research?
Skipping the verification step. Most kdp keyword research problems are caught by a 10-minute pre-flight check before upload — we see this in our formatting queue every week.
How much time does kdp keyword research usually take?
Allow 2-8 hours for a first attempt, 30-60 minutes once you've done it twice. The first time eats time because you're learning the controls; subsequent times are mechanical.
Are the free tools good enough or should I pay?
Free tools work if you have time to learn them. Paid tools (or services) save 10-30 hours and reduce rejection rates. Worth it if you're launching multiple titles.
Where can I check my work before going live?
Run a free KDP Readiness Score — catches 35+ common issues in 60 seconds, no signup. If anything fails, the report tells you exactly what to fix.
About this guide
Written by Robert Prime for publishing.co.uk. Last reviewed May 2026. Specs and pricing change — verify current figures with the linked sources before relying on them.
