How to choose your seven KDP keywords

Choose your seven KDP keywords by covering seven distinct ways a real reader might look for your book. Use relevant short phrases, remove phrases already obvious from the title and other metadata, and avoid misleading, promotional, or unrelated terms. The goal is an accurate map of reader intent, not the highest possible search count.

What does a KDP keyword field do?

KDP provides up to seven keyword fields for words or short phrases that describe a book. Amazon's official KDP keyword guidance emphasizes relevance and identifies terms to avoid. Read the current page before uploading because platform rules can change.

Keywords are only one part of discovery. The title, subtitle, description, categories, cover, book content, and reader response all provide context. A keyword cannot repair an unclear book concept, and entering a phrase does not guarantee that a book will rank for it.

How do you build a useful keyword list?

1. Write the book's core facts. Note the subject, reader, problem, desired outcome, approach, setting, and format.
2. Turn those facts into reader language. Ask what someone would type before they knew your title. Prefer ordinary phrases over internal publishing labels.
3. Collect several sources. Review retailer suggestions, relevant search results, public reader questions, and the wording used by directly comparable books. Treat these as idea sources, not permission to copy protected phrases or metadata.
4. Group phrases by intent. Put audience phrases together, problem phrases together, and method phrases together. This reveals repetition.
5. Check relevance against the manuscript. A phrase belongs only if a reader using it would reasonably expect your book.
6. Remove prohibited or risky terms. Check KDP's live rules for misleading claims, unrelated names, promotional wording, and other restrictions.
7. Choose seven distinct fields. Give each field a job and favor coverage over minor word variations.
8. Save the reasoning. Record why each phrase fits and the date you checked it. This makes later updates deliberate.

Breakthrough's Metadata Keywords tool can suggest phrases for the seven KDP fields. Starter adds 5 Metadata Keywords credits each month, Plus adds 20, and Elite has unlimited use. Each metered run uses one credit; unused credits roll over and the next allocation is added on the renewal day. Suggestions still need a relevance and policy check.

What seven kinds of intent can you cover?

There is no universal seven-part formula, but this worksheet produces variety.

These are invented examples that demonstrate intent types. They are not researched recommendations for a particular book.

You do not have to use every role. A history book, workbook, memoir, or technical manual will produce different search behavior. The useful principle is that seven near-identical phrases usually cover less ground than seven accurate, meaningfully different phrases.

How do you find the language readers use?

Start with your topic-research notes. If you have not yet validated the idea, follow how to choose a book topic with real demand first. Keyword work should refine a known reader need, not create an artificial market after the manuscript is done.

Search the retailer as a reader would. Type the beginning of a natural phrase and observe suggestions, but do not treat a suggestion as proof of volume. Open genuinely relevant results. Note recurring nouns, problems, reader descriptions, and formats. Then check broader web searches and public questions for alternative wording.

Keep an evidence table:

Writing “weak” is more useful than forcing every candidate into the final seven.

Should you use broad or specific phrases?

Broad phrases may describe a large field but say little about why your book is relevant. Very specific phrases may be accurate but use wording no reader naturally chooses. Mix levels only when each phrase reflects real intent.

For an invented book about growing herbs on apartment windowsills, “gardening” is accurate but extremely broad. “Indoor herb gardening for apartments” is clearer about subject and situation. “Northern-facing fourth-floor kitchen basil solutions” may be too peculiar unless the whole book genuinely focuses on that constraint and readers use similar language.

Read every candidate as an expectation: if someone searches this phrase and opens the book, will the description and table of contents confirm the match? If not, remove it.

What repetition should you remove?

Check repetition across the whole metadata package, not only the seven fields. Your title and subtitle should already state the main promise. The description explains value in natural prose. Categories indicate shelving. Keyword fields can add relevant language those elements do not cover well.

Remove:

• singular and plural versions with no meaningful difference;
• seven reordered versions of the same two or three words;
• a full title copied into a keyword field;
• the author name when it adds no legitimate discovery context;
• category labels repeated mechanically;
• spelling variants nobody in the intended market is likely to use.

Do not sacrifice clarity merely to avoid one repeated word. If “cookbook” naturally belongs in two distinct phrases with different intent, relevance matters more than an artificial no-repeat rule.

Which keywords should you avoid?

Do not use a term simply because it appears popular. Avoid wording that misrepresents the book, implies an unsupported association, uses another author or brand improperly, makes unverifiable claims, or promises content the manuscript does not contain. Do not use metadata to advertise price, rank, reviews, or temporary promotions.

KDP's metadata rules also apply to the larger listing. Review the official metadata guidelines and ensure the information on the cover and in the title fields is consistent.

How do you make the final choice?

Score each phrase with three plain questions:

1. Relevance: Is this genuinely central to the book?
2. Reader language: Is this a plausible way the intended reader describes the need?
3. Added coverage: Does it add a meaning the other fields and visible metadata do not already cover?

Choose the strongest balanced set, then read the seven together. They should describe one coherent book from several angles, not seven different books. Save the unused candidates and revisit the fields when positioning changes or when current evidence shows that readers use different language. Keyword selection is careful metadata work, not a one-time trick.