Keywords decide whether Amazon ever shows your book to anyone. They cost nothing to change and nothing to get right — which makes it strange how many KDP books ship with seven empty backend boxes, or worse, seven slots of guesswork copied from a YouTube video.
This is the whole process: where Amazon reads keywords from, what the 7 backend slots are actually for, the words that can get a listing suppressed, and how to check a keyword has real demand before you build a book on it.
Where Amazon reads keywords from
Your book is searchable through four channels, in roughly this order of weight:
- Title and subtitle. The strongest signal by far. A phrase in your title ranks you for that phrase more powerfully than the same phrase anywhere else.
- The 7 backend keyword slots. Set in the KDP wizard, invisible to shoppers. Each box takes up to 50 characters.
- Categories. Not keywords strictly, but they define which browse shelves and “#1 in X” badges you can appear on.
- The description — Amazon has indexed it for search since 2023-ish, weakly. Write it for humans, not for the algorithm.
The practical consequence: your single most important keyword decision is the title/subtitle, not the backend. The backend slots are where everything that didn’t fit goes.
How the 7 backend slots work
Rules that follow from Amazon’s own keyword documentation:
- Phrases beat single words. Each box takes a phrase up to 50 characters. “anxiety journal for teen girls” covers “anxiety journal”, “teen anxiety”, “journal for girls” — the words in one box combine and match in any order.
- Because words combine, don’t waste slots on reorderings. “dog training book” and “book dog training” are the same entry.
- Don’t repeat what’s already in your title or subtitle. Those words are indexed. Backend slots repeating them are wasted inventory.
- No quotation marks, no need for commas between variants inside a box, no semicolon tricks. Plain lowercase phrases.
- Singular vs. plural: matching handles most of it. Spend slots on genuinely different phrases instead.
Seven slots × 50 characters is enough for 15–25 distinct phrases if you pack them as combinable word groups. Most books use a fraction of that.
Words that get listings suppressed
Amazon’s metadata rules prohibit specific classes of words in titles and keywords, and enforcement got stricter over the last couple of years. From Amazon’s own guidelines, do not use:
- “free”, and price or promotion claims of any kind
- “bestselling”, “best seller”, or any sales-rank claim
- Subjective quality claims — “best”, “greatest”, “amazing”
- “new” or anything only temporarily true
- Other authors’ names, other books’ titles, or brand names you don’t own — including in backend slots where you think nobody sees them
- Anything misleading about the book’s content
The penalty ranges from the keyword being ignored to the listing being suppressed from search entirely — and you are not notified when it happens. We learned this list the careful way: KDPie’s listing builder had to encode these rules as hard filters, including the German equivalents (kostenlos, gratis) for Amazon.de publishers, because generated metadata that trips them is worse than none.
Finding keywords worth researching
Start where the demand is recorded:
Amazon autocomplete. Type your seed phrase into Amazon’s search bar (in the Books department) and harvest the suggestions, then add a letter — “dog training a…”, “dog training b…” — and harvest again. These are real queries from real shoppers, ordered roughly by popularity. Fifteen minutes gives you 30–60 candidates.
The shelves of adjacent books. Open the top books for your seed phrase and read their titles and subtitles: the phrases they lead with are the phrases someone already validated with money.
Validating demand — the step everyone skips
An autocomplete suggestion tells you a phrase is searched, and nothing about how much or by whom. Two honest ways to find out:
The free way: search the phrase, pull the Best Sellers Ranks of the top 10 results, and judge whether the shelf actually sells — our BSR guide walks through the thresholds, and the BSR calculator converts ranks to sales ranges. If the “top” books for a phrase idle beyond rank 500,000, shoppers may type the phrase, but they aren’t buying from it.
The measured way: tools with Amazon search-volume estimates give you the demand number directly. One warning that applies to all of them, ours included: Google volume is not Amazon volume. A phrase can be huge on Google as an information query and near-zero on Amazon as a purchase query — gift-shopping phrases skew the opposite way. If a tool shows one number and doesn’t say which engine it came from, assume Google, and trust it less. (For what each tool can and can’t measure, see our comparison of the research tools.)
Whichever way you validate, check the other side of the equation too: demand with thirty strong incumbents is not an opportunity. Demand plus a weak or aging shelf is. That two-sided judgment — demand and contestability weighed together — is exactly what KDPie compresses into a per-niche verdict, with the top-shelf evidence and a generated 7-slot backend set included. The $7 trial covers enough scans to test it on your own candidates.
A note on low-content books
For journals, planners and logbooks the mechanics are identical, with the audience attached to the phrase doing most of the work. “Blood pressure log large print” and “blood pressure log” are different niches with different buyers and different shelves. Low-content shelves also saturate fast: the specific phrase with a real audience and a thin shelf is worth ten generic ones. Validate before designing — the design is the cheap part.
The 20-minute checklist
- Harvest 30+ candidates from autocomplete and competitor titles.
- Validate the 5–8 most promising ones against the shelf (BSRs, publication dates) or measured Amazon volume.
- Put the strongest phrase in the title/subtitle. Keep it readable — a keyword-stuffed subtitle repels the human who has to click.
- Pack the runners-up into the 7 backend slots as combinable phrases, no repeats of title words.
- Sweep for banned words: free, bestselling, best, new, brands, other authors.
- Revisit in a month: books whose shelves you validated are worth re-checking as they age.
None of this requires paying anyone. The paid tools compress the hours and add the data you can’t see by hand — measured demand and rank history. The method stays the same either way.