You can find KDP keyword ideas by combining your manuscript’s specific promises with the language readers use in Amazon search, comparable books, reviews, and reading communities. You do not need a paid keyword subscription to build a useful shortlist. The trade-off is manual research time and less information about demand.
These eight methods help you gather candidates before choosing the keywords for your book. Amazon permits up to seven keywords or short phrases; eight research methods do not mean eight keyword fields.
The examples use a fictional cozy mystery, Lanterns at Maple Quay: a retired librarian investigates a missing historical map in a coastal village. It features community friendships, no graphic violence, and no central romance or supernatural plot. The sample phrases illustrate decisions; they are not measured searches or sales results.
1. Mine the manuscript for specific reader promises
Start with what the book delivers, rather than what looks popular. List its setting, character types, plot themes, and tone. For nonfiction, use the reader’s problem, intended audience, and the practical outcome the book genuinely supports.
Amazon identifies setting, characters, themes, and tone as useful keyword types. Turn those details into possible reader phrases:
| Detail in the fictional manuscript | Candidate to investigate | What must be true |
|---|---|---|
| A retired librarian investigates | retired librarian sleuth | The librarian has a central investigative role. |
| The story takes place in a coastal village | coastal village secrets | The setting matters to the story. |
| A historical map goes missing | missing historical map | This is a central plot element, not a passing reference. |
| The mystery has no graphic violence | gentle amateur detective | The tone matches the expectation the phrase creates. |
These are starting hypotheses. A truthful description is not automatically a phrase readers search for.
Use the same intent-matching principle that guides useful web content: the result must satisfy the person’s actual task. For a book, that includes the reading experience. A romance phrase would mislead readers if romance is only incidental. Keep that mismatch out of the candidate list even if another author uses it successfully.
2. Collect suggestions from Amazon’s book search
Choose the storefront and language your readers use, then search within Books or the Kindle Store, as appropriate for the format you are researching. Amazon explains that searching within these departments produces more relevant book results than searching across all products.
Enter the beginning of a seed phrase and note relevant completions. For the fictional mystery, prefixes to investigate might include “librarian mystery” or “coastal mystery.” Those are example inputs, not suggestions observed during research.
Record the typed prefix, exact suggested phrase, department, storefront, language, and date. Then search the complete phrase and inspect whether the results fit your book.
Do not label the first suggestion the highest-volume keyword or a missing suggestion “zero demand.” A completion is a lead to investigate; it does not supply a monthly search count or prove that your book can rank.
A private window can reduce the influence of your own browsing history, but it does not make every result neutral or identical across locations. Keep the research context consistent instead of treating one screenshot as a universal ranking.
3. Study the promises of comparable books
Search your strongest seeds and select a small set of books that serve the same audience, genre, and reading experience. Start with five as a manageable sample, expanding it when the results mix different kinds of books. Five is a research choice, not a KDP rule.
Separate sponsored placements from organic results. For each comparable, record its product URL or ASIN, format, visible title and subtitle, relevant description wording, and the observation date. Read the sample where available to check whether the marketing promise matches the content.
Look for descriptors shared by genuinely comparable books. If several descriptions emphasize an amateur investigator and a close-knit community, those ideas may deserve investigation. Their appearance does not prove that they are hidden keywords, high-volume searches, or the cause of sales.
Keep author and title names in your research references, not your keyword fields. Amazon’s metadata guidelines prohibit references to other authors and their books as keywords. Adding “books like” does not provide a supported exception.
The useful output is a set of relevant reader promises you can express in your own accurate language, with a traceable source for each idea.
4. Read reviews for recurring expectations
Reviews can reveal how readers describe a book after experiencing it. Read a mix of positive and critical reviews for your book, if available, and for close comparables.
Collect recurring expectations rather than isolated adjectives. A pattern of praise for an easygoing investigation may suggest researching gentle-mystery language. Complaints that a supposedly light story contains graphic violence reveal an expectation mismatch, not a keyword to copy.
Use a simple note structure:
Book URL or ASIN:
Review date / source link:
Reader expectation, paraphrased:
Supported by our manuscript? yes / no / uncertain
Candidate phrase:
Further store check needed:
This is a template, not a record of a review we observed. Preserve the original reference so another team member can check the interpretation. Avoid turning paraphrases into invented quotations.
Review language does not equal search language. A reviewer may describe a book as “beautifully written” without ever searching that phrase. It can help clarify your description, but subjective praise is not a suitable keyword strategy. Keep only specific expectations your book fulfills, and send any resulting candidate back through an Amazon search check.
5. Browse category pages for adjacent vocabulary
Browse the relevant genre and subgenre pages in the target storefront. Note how the shelves divide reading experiences, then inspect the books within the closest matches. This can reveal a more precise audience or theme that a broad seed missed.
For the fictional mystery, the useful question is whether the book belongs alongside gentle amateur investigations, not whether an unrelated category appears easier to enter. Record the category path and the specific descriptor you want to investigate.
Amazon currently lets authors select three categories in KDP. Its category guidance explains that available categories differ by marketplace and format, and selections should accurately describe the book.
Category browsing supplies context; it does not measure keyword demand. Nor does a visible category prove that one exact phrase will unlock it. Amazon says keywords can help determine shelving, but that is not a guarantee of placement or a bestseller badge.
Before using a discovered term in the keyword fields, compare it with your selected categories. Amazon advises avoiding words already covered by those categories. Keep the shelf information in the research sheet and use the keyword space for relevant details that add information.
Use category browsing to refine the book’s audience and identify specific descriptors worth investigating.
6. Read recommendation requests in reader communities
Browse public book-discussion communities for requests about the reading experience your manuscript provides. Recommendation requests are useful because they describe what someone wants before choosing a book, while reviews describe the experience afterward.
Look for three parts of a request: the desired experience, the audience or setting, and explicit exclusions. For example, a hypothetical request for a small-town investigation without graphic violence points to a different promise from a dark crime thriller.
Record the thread URL, date, and a short paraphrase. Then ask:
- Which part of the request does the manuscript actually fulfill?
- Which exclusion would make it a poor recommendation?
- Can the relevant idea become a clear, non-branded search phrase?
Do not copy a recommended author’s name into your KDP keywords. Extract the underlying preference instead. A reader asking for “more like” a specific writer may mean tone, pacing, setting, or character type; those are different candidate directions.
A busy thread is not a count of Amazon searches. Treat community language as a hypothesis source and validate it in the appropriate storefront. You can do this research by reading existing public discussions; posting promotional replies or contacting participants is not necessary.
7. Use web questions to find alternative phrasing
Search the broader web for questions about your genre, topic, or reading experience. Google autocomplete and relevant articles can expose unfamiliar wording that is absent from your initial manuscript notes.
Try a seed plus a reader need, such as “mystery with” or “gentle detective stories.” For nonfiction, the equivalent might be a specific beginner problem rather than the broad subject. Save useful wording with its actual source, not as an assumed Amazon observation.
This method is most helpful when it reveals a synonym, a reader’s problem, or a different way to describe the same experience. It is less useful when it produces a general informational query with no clear connection to choosing your book.
Keep the data sources separate. Google suggestions, Google Trends interest, and website keyword volumes do not measure Amazon book searches. Screpy’s guide to separating keyword data sources explains why source and market context matter.
Take promising wording back to Amazon’s Books or Kindle Store and inspect the result type. If a phrase produces unrelated products or a different kind of book, revise or reject it. Web research expands vocabulary; it does not replace store validation.
8. Ask early readers how they would search
If you already have beta readers or an author audience, ask for unprompted wording before showing them your candidate list. This can reveal a mismatch between how you describe the book and how its intended readers would seek it.
Use a short prompt like this:
After reading the sample or manuscript:
1. What kind of reading experience would you recommend it for?
2. What would you type into a bookstore search to find something like it?
3. What wording would make you expect a different book?
Please answer before looking at my proposed keywords.
Record whether the person read a sample or the whole book, and keep the response linked to an internal research reference. Do not include private reader details in a public keyword sheet.
Compare the answers with the manuscript and with public store results. If someone suggests romance language for the fictional mystery, investigate why the sample created that expectation before using it.
A small, friendly beta-reader group is not representative of the whole market. People’s stated search ideas may also differ from what they actually type. Use the answers to generate candidates and detect confusion, not to declare search volume or conversion potential.
If you have no early readers, leave this source unused and work with the public methods above. There is no need to buy responses or delay all keyword research until an audience exists.
Turn candidate phrases into a defensible shortlist
Combine the ideas in one sheet, preserving their sources. Deduplicate phrases without deleting the evidence that led to them. For every remaining candidate, check manuscript fit, relevant store results, and compliance with Amazon’s keyword rules.
Use keep, hold, and reject decisions. Keep a phrase only after the required checks support it; hold it while evidence is missing; reject a misleading or prohibited term regardless of apparent popularity.
Here is a copyable CSV for the fictional book. None of these rows represents an observed search:
id,phrase,source,source_reference,market,manuscript_fit,store_check,demand,decision,reason
K01,retired librarian sleuth,manuscript,protagonist outline,Amazon.com,yes,not observed,unknown,hold,check results before selection
K02,coastal village secrets,manuscript,setting notes,Amazon.com,yes,not observed,unknown,hold,confirm reader intent
K03,missing historical map,manuscript,central plot,Amazon.com,yes,not observed,unknown,hold,check whether results match fiction
K04,paranormal romance,brainstorm,unsupported idea,Amazon.com,no,not observed,unknown,reject,book has neither central romance nor supernatural plot
Add dated URLs or screenshots when you perform the checks. Prefer a smaller verified shortlist to filling the sheet with guessed metrics.
Do not convert Amazon result counts into monthly demand: listed products measure a different thing from searches. Likewise, Best Sellers Rank is not a keyword’s search volume or your position for a query. Amazon describes sales rank as relative customer activity, with separate ranks for individual book formats.
You can still make a useful relevance decision when demand is unknown. What you cannot honestly conclude is that the phrase has a particular search volume, a guaranteed competition level, or proven profitability.
Prepare your KDP fields and record the update
After validation, choose up to seven accurate keywords or short phrases. Amazon’s metadata guidelines recommend phrases of two or three words. Treat that as guidance toward clear, specific wording, not a promise that one fixed length ranks best.
Remove information already covered by the title, contributors, and selected categories. Follow the live character counter in each KDP field. Do not cram unrelated words together to fill every available space.
If the first three candidates in the example later pass store validation, a draft might look like this:
| Keyword field | Illustrative draft |
|---|---|
| 1 | retired librarian sleuth |
| 2 | coastal village secrets |
| 3 | missing historical map |
| 4–7 | Leave unused until further relevant candidates are validated. |
This is a field-planning example, not a tested keyword set. The phrases are not enclosed in quotation marks. Recheck the final draft against Amazon’s keyword guidance, including restrictions on other authors, book titles, unauthorized brands, promotions, and misleading claims.
Keep the actual book title consistent with the cover instead of turning it into a keyword list. Record any words removed because they already appear in the title or categories.
To update keywords, Amazon directs authors to Bookshelf → the book’s ellipsis menu → Edit book details for the relevant format. Edit Keywords, choose Save and Continue, and finish by selecting Publish on the Pricing page. Saving an intermediate page does not complete that workflow.
Save the old and new field values, the date submitted, the format, and the date the update was verified live. Consult Amazon’s current publishing timelines for the applicable change rather than assuming an instant update.
Review comparable periods afterward and record other changes, such as ads, price promotions, or a new cover. Orders and sales-rank movements alone do not show which organic keyword caused a sale. Keep query observations and sales outcomes distinct, then revise the shortlist when stronger evidence becomes available.