Keywords for books come in three kinds, and most bad keyword advice comes from mixing them up. Store metadata keywords are the up to seven keywords or short phrases in your KDP listing, advertising keywords are phrases you bid on, and discovery keywords are the words readers use to describe the book they want, which in fiction are usually trope names.
Choose KDP’s up to seven keywords from the words readers actually use, not search volume alone. Our rule of thumb for filling all seven: three on your strongest trope stack, two on subgenre and setting, one on character type or tone, one left flexible. Then check each phrase against the books already on the shelf.
These are the words in your listing — on Amazon KDP, up to seven keywords or short phrases, and KDP’s own help says single words work better than phrases and specific words work better than general ones. They cost nothing, they are editable at any time, and they are one of the few levers you fully control after publication. Choose them for precision: a keyword or short phrase you can realistically rank for and that your book genuinely satisfies, because Amazon rewards clicks that turn into reads. The detailed mechanics live on our KDP keywords guide.
Phrases you bid on inside an ad platform. Different economics entirely: here a broad, expensive term can still be worth it if the conversion holds, and a cheap term can be worthless if it brings the wrong reader. This is the one place where a search-volume and competition tool genuinely is the right instrument — Publisher Rocket and its alternatives exist for exactly this job, and we do not compete with them on it.
This is the layer most guides skip. When a reader describes what they want, in a review or a BookTok comment or a forum thread, they do not use category names. They say “enemies to lovers but he falls first”, “forced proximity, one bed, no third-act breakup”. That vocabulary is trope vocabulary, and it is the vocabulary your blurb, your covers, your ad copy and your metadata should all be speaking. We have analyzed 2,860,000+ demand signals, 113,500+ of them from BookTok comments and captions; each figure was counted 2026-11-05 10:15 UTC.
Choosing keywords on volume alone. A volume tool cannot tell you how often readers mention a trope. Demand-signal counts show that, and comp titles show how crowded the shelf already is. Use all three: volume, demand signals and comps.
Demand signals per trope, how many came from BookTok comments and captions, and the share tagged underserved or novel by Tropesmith's classifier. A high count means readers mention the phrase often. The tag is the classifier's call on each text alone, so check the comp titles before you spend a keyword slot on it.
| Trope phrase | Demand signals | From BookTok | Tagged underserved or novel |
|---|---|---|---|
| found family | 300,602 | 9,371 | 45% |
| slow burn | 125,290 | 9,650 | 33% |
| forced proximity | 125,150 | 6,247 | 33% |
| second chance romance | 71,239 | 748 | 36% |
| enemies to lovers | 59,379 | 6,844 | 32% |
| dark romance | 40,738 | 6,340 | 49% |
| reverse harem | 21,209 | 2,773 | 60% |
| hurt/comfort | 17,417 | 7,325 | 46% |
| age gap | 12,771 | 2,075 | 43% |
| grumpy sunshine | 12,570 | 2,908 | 27% |
| fake dating | 10,468 | 2,008 | 27% |
| he falls first | 6,128 | 2,556 | 34% |
Counted 2026-10-07 17:24 UTC. A demand signal is one specific thing a reader asked for or praised in a review, comment or post, picked out by Tropesmith's classifier. One review can give several. Each signal gets a rarity tag (common, underserved or novel) from the text of that review or post alone; the classifier is never shown what is on the shelf. A tag points at where to look. It does not prove that no book answers the signal. Shares are the tagged signals divided by all signals in the same scope. How we count. Each row counts every demand signal that mentions the phrase's trope as a main or secondary tag, so rows overlap; “From BookTok” counts signals from BookTok comments and captions. These are demand figures, not Amazon search volumes — we do not have search volumes and do not claim to.
The corpus these demand numbers are counted from.
Counted 2026-11-05 10:15 UTC by direct query against the Tropesmith corpus. Method: how we count demand signals · how a Map is built.
A Tropesmith Map scores your exact trope stack against demand and supply data — and shows where to look next.
Build my Map → See a real sample Map first → See pricing →Choose KDP’s up to seven keywords from the words readers actually use, not search volume alone. Our rule of thumb for filling all seven: three on your strongest trope stack, two on subgenre and setting, one on character type or tone, one left flexible. Then check each phrase against the books already on the shelf.
Book keywords are the phrases attached to a book so that a store, a search engine or an ad platform can decide who to show it to. There are three distinct kinds and they are often confused: store metadata keywords (the up to seven keyword slots Amazon KDP gives you, which shape which searches and also-boughts you appear in), advertising keywords (phrases you bid on, which cost money per click), and discovery keywords (the words readers actually use when they describe what they want, which are usually trope names). They are chosen differently and confusing them is the most common and most expensive mistake.
On Amazon KDP, up to seven keywords or short phrases; KDP's help says single words work better than phrases and specific words work better than general ones. Words already in your title and subtitle are indexed anyway, so repeating them wastes a slot. Advertising keyword lists are not capped the same way; there you are limited by budget and by how many phrases you can actually manage.
Start from what your book genuinely delivers, gather candidate phrases from Amazon's own autocomplete, and then choose between the finalists on demand signals and comp titles as well as volume. Tropesmith tags the 2,591,893 demand signals counted on 2026-10-07 with the tropes they mention and counts how many mention each one, which shows how often readers mention a trope.
For fiction, they are often the best ones available, because trope names are how readers actually describe what they want to each other. Someone hunting for a book on BookTok or in a Goodreads review rarely says 'contemporary romance' - they say 'enemies to lovers, forced proximity, he falls first'. We have analyzed 105,485 demand signals from BookTok comments and captions alone (counted on 2026-10-07).
Yes, and it is worth doing before paying for anything. Amazon's search autocomplete shows real phrases shoppers type. Also-bought rows show adjacency. And the trope demand pages and lane snapshots on this site are free to read with no signup. Paid keyword tools add search-volume estimates and competition scoring on top of that.
Demand figures on this page come from the Tropesmith demand-signal engine: per-trope counts and tagged shares counted 2026-10-07 17:24 UTC. This guide is general publishing information and is not affiliated with Amazon. Part of the market intelligence hub.