The first time I watched an experienced SEO set up a Kindle book listing, they did the sensible thing. They opened a keyword tool, sorted by monthly search volume, took the highest-volume phrases that were not hopelessly competitive, and pasted them into Amazon’s seven keyword fields. Then they wrote a subtitle containing three of those phrases, because why waste the space.
Every one of those moves is correct on Google. On Amazon, four of the seven fields were wasted on words that were already in the title, the subtitle broke Amazon’s own metadata rules, and the phrases described what people read about on the open web rather than what people reach for in a shop.
Both platforms have a search engine, an index, and a ranking system, which is why Amazon SEO can look deceptively similar to Google SEO. That surface similarity is exactly what causes the damage, because it invites you to carry across a method built to answer a completely different question.
Google answers a question, Amazon fills a basket
Start with the person typing, because everything else follows from their state of mind.
A Google searcher is unresolved. They want to know something, compare two things, find a place, or decide whether a problem is worth solving. Most of the volume on Google is informational, which is why content marketing works there at all. You can catch someone early, be useful, and hope to sell to them later.
An Amazon searcher has already decided to buy. They have opened a shop. The only open question is which item they carry to the checkout. There is no awareness stage, no consideration funnel that you get to nurture, no informational traffic to build on. The listing is the entire funnel.
That changes what the word “relevant” means. On Google, a relevant result answers the question. On Amazon, a relevant result is one that a person in a buying mood will click and then pay for. Those are not the same test, and a keyword can pass one while failing the other badly.
Here is the clearest example I know. “How to start journaling” is a perfectly good Google target and deserves an article. On Amazon it is close to worthless, because it describes a question rather than an object. The shopper equivalent is “guided journal for beginners”. Same topic, same person, completely different words, because in one case they want an explanation and in the other they want a thing with a cover on it.
If you take nothing else from this piece: Google keyword research collects questions, Amazon keyword research collects products.
What Google SEO and Amazon SEO Actually Index
The second big difference is how much room you have.
On Google, your keyword surface is effectively unlimited. Title tag, headings, body copy, image alt text, internal links, schema, plus everything off-page. If you find a term you want, you can write another page and go after it. Keyword research on Google produces a content plan, and the plan can always get bigger.
On Amazon, a book has a small fixed set of fields, most of which are set at publication and tied to the physical cover:
- Title, which Amazon’s metadata guidelines say should contain only the actual title as it appears on the cover
- Subtitle, held to the same rules, with title and subtitle together required to come in under 200 characters
- Series name, which takes the series name only
- Contributor names
- Seven keyword fields, which shoppers never see
- Three categories, chosen at setup
- Description, which does far more work as a sales asset than as a ranking asset
That is the whole board. You cannot add a page. You cannot write your way to more coverage. Amazon keyword research does not produce a plan, it produces a packing problem, and the skill involved is allocation rather than production.
Which makes one line on Amazon’s own keyword help page the most valuable sentence in KDP metadata, and the one almost nobody acts on. Its list of what not to put in the keyword fields includes information already in your title, contributors, or categories. You are already indexed for those words. Every one you repeat is a slot spent buying something you owned.
The same page says something else a Google-trained marketer will find strange: single words work better than phrases, and specific words work better than general ones. It explains that wrapping a phrase in quotation marks narrows who can find you, whereas listing the individual words lets shoppers searching any of them reach the book. Amazon is asking you for vocabulary. Google asks you for phrases. That distinction quietly breaks most imported workflows.
The comparison in one table
The differences become clearer when you put Google SEO and Amazon SEO side by side.
| Amazon KDP | ||
|---|---|---|
| What the searcher wants | An answer | An object to buy |
| Dominant intent | Informational | Transactional |
| Where your keywords live | The whole page, plus off-page signals | Title, subtitle, series, contributors, seven keyword fields, three categories |
| Surface area | Unlimited, you can always add more | Fixed, and mostly locked to the cover |
| Official volume data | Yes, from the ads platform | None published |
| How matching works | You target phrases | Specific individual words, recombined by Amazon |
| Off-page signals | Links move rankings | Links move traffic, not position inside Amazon |
| The metric that counts | The click | The purchase |
| Feedback loop | Weeks to months | Days, via sales and advertising search term reports |
| Revised later? | Continuously | Editable, but most people never revise |
Two rows in that table cause most of the trouble. The volume row, and the metric row. They are connected.
Why Google Search Volume Doesn’t Work the Same Way on Amazon
The Google-trained reflex is to sort by volume and work down. On Amazon that reflex fails three times over.
First, there is no official Amazon search volume. Amazon does not publish it. Every number you have ever seen in a KDP research tool is modelled by a third party from scraping and inference. Some of those models are decent. None of them are ground truth, and different tools will rank the same phrase differently. Treat any volume figure as a rough ordering, never as a measurement.
Second, even a perfectly accurate volume number would not answer the question you actually have. Amazon demand is already narrow and already commercial. A phrase with modest traffic and four weak books on the first page beats a phrase with heavy traffic and forty entrenched ones, and volume alone cannot tell those apart.
Third, volume is not distributed the way you expect. Someone typing a broad category name is at the start of a browse and will scroll, filter and change their mind. Someone typing a five-word phrase is often one click from paying. On a platform whose ranking rewards selling, conversion is not just the goal. It is the input.
So what should you look at instead? A good KDP keyword research process starts with four things, in this order.
Competition depth, not competition count. Ignore the number of results. Open the first page and read the top twenty listings as a buyer would. Do the covers look professionally made or homemade? Do the leaders have review counts accumulated steadily over years, or a small burst from one launch? And most usefully: how far down does relevance survive? If results twelve through twenty are only loosely related to the query, Amazon is padding the page because it does not have twenty good matches. That is not a crowded keyword. That is an under-served one, and it is your opening.
Buyer intent inside the words themselves. Product-shaped language beats question-shaped language every time. Words like workbook, guide, planner, for beginners, large print, with practice questions, second edition, and pocket all describe an object someone can picture holding. Phrases beginning with why does, how do I, or what is describe someone who walked into the wrong shop. Sorting on this alone takes five minutes and removes more junk than any volume filter.
Category rank thresholds. Amazon lets you pick three categories, and each one has an implicit price of admission measured in sustained sales. The price varies enormously between categories that look similar in the dropdown. The check is simple: open the bestseller list for a candidate category, look at the Best Sellers Rank of the book sitting at position twenty, and ask honestly whether your launch can hold it. If not, you do not have a keyword problem. You have a category problem no keyword work will fix.
Price band. Shoppers compare on price faster than they do anything else, and the results page for a phrase sets an expectation before you get a word in. If every top result sits at one price point and you are at three times that, the phrase is not really available to you unless the cover visibly justifies the gap. Keyword and price are one decision here, not two.
The awkward part of all four checks is that they are downstream of a candidate list you do not have yet. You need phrases before you can assess them, and mining your own manuscript is a poor way to produce them, because you will generate the words an author uses rather than the words a shopper uses. This is the one step where a keyword generator can save time. Feed it a plain description of the book and use the resulting phrases as a starting pool, not as a source of authoritative demand data. Read those demand figures as ordering rather than truth, for the reason above.
What matters more is the breadth of the list, and whether suggested categories arrive in the same pass as the keywords, because on Amazon those two choices are a single decision. Doing them separately is how people end up with a genuinely good keyword set pointed at a category they were never going to place in. That is why keyword research and category selection should be evaluated together rather than treated as separate tasks.
Why Long-Tail Keywords Work Differently on Google and Amazon
Everyone agrees long tail is good. Almost nobody notices that it is good for different reasons on each platform.
On Google, long tail is an accumulation strategy. Each specific query is small, but you can build a page for every one of them, and the total adds up. The cost of chasing one more long tail term is one more piece of content, and the strategy scales as long as you can keep publishing.
On Amazon you have one product. You cannot spawn a listing per phrase, so nothing accumulates. What long tail buys you here is match precision: fewer impressions, far higher conversion, and a ranking system that reads those conversions as evidence you deserve the page. A handful of well-matched sales moves you further than a pile of impressions from a broad term nobody clicks.
There is a second difference, and it is the one that changes how you actually fill the fields. On Google, a long tail keyword is a string you target as a string. On Amazon it is more useful to think in components, because the keyword fields are indexed as words that Amazon recombines against queries. The way to cover “large print sudoku puzzle book for seniors” is not to type that sentence into a slot. It is to make sure the component words, minus the ones already sitting in your title and category, are present across your fields. You are stocking a vocabulary, not registering queries.
And there is a floor. Below a certain level of specificity, an Amazon phrase has no shoppers at all, and unlike Google there is no residual informational traffic keeping it alive. A phrase nobody types is worth zero, and you cannot tell which of your beautifully specific candidates are real until you have data. That is what the advertising step below is for.
5 Google habits that break a KDP listing
1. Stuffing the subtitle. On Google, extra relevant words in the title tag are roughly neutral to positive, so people carry the habit over. Amazon’s metadata guidelines treat title and subtitle as descriptive fields that must match the cover, cap the pair under 200 characters, and prohibit repeating generic keywords, referencing sales rank, and promotional wording. There is enforcement risk and there was never much upside, because the seven hidden fields exist to carry what the cover cannot say.
2. Chasing head terms. Cookbook. Romance. Self help. On Google a head term is a stretch goal you build toward over a couple of years, and the intermediate positions still pay you something. On Amazon a head term is a page owned by books with long sales histories, and there is no partial credit for ranking eightieth. Nobody scrolls to eightieth in a shop.
3. Copying competitor titles and keywords. Pulling a rival’s ranking terms and targeting the overlap is standard on Google and close to useless here. A bestselling book ranks for its terms because it sells, not the other way around, so the terms are a symptom rather than a cause. Copying them puts you on the exact pages where you are least equipped to compete.
4. Treating categories as filing rather than strategy. Most people pick their three categories last, in a hurry, on publication day, then wonder why the keyword work did nothing. Amazon uses categories and keywords together to decide which lists and result pages you belong on, and its category guidance tells you to use keywords for the specificity the category tree does not cover. Picking three broad, over-subscribed categories and hoping keywords will rescue you is backwards. Pick the category you can actually place in, then spend the fields on everything it does not already say.
5. Filling the fields once and never returning. This is not an imported habit, it is the absence of one. On Google you check Search Console and iterate forever. On Amazon most people fill the seven fields the night before launch, with no data at all, and never touch them again. The fields are editable at any time, and running even a small automatic sponsored products campaign produces a search term report showing the actual phrases shoppers used to reach your book. That is the closest thing to Search Console that Amazon will give you, and it costs very little to obtain.
A routine you can run tomorrow
- Write the book’s promise in one plain sentence, phrased the way a shopper would say it out loud, not the way you would describe it at a party.
- Mine autocomplete. Set the department to the store your book sells in, type your core noun, and record every completion. Then type each completion plus one more letter and record those. Twenty tedious minutes, and the only first-party demand signal Amazon gives away.
- Generate a wider candidate set from a plain description of the book, so you are not trapped inside your own vocabulary.
- Delete everything question-shaped and everything that names a topic rather than an object.
- Take your top ten survivors and open the results page for each. Note cover quality, review depth, price band, and the position at which relevance starts to decay. Keep the phrases where relevance decays early.
- Choose your three categories from what those pages showed you, not from the taxonomy. Check the Best Sellers Rank of the book at position twenty in each candidate list and be honest about whether you can hold it.
- Write out every word already present in your title, subtitle, series name and author name, and cross all of them off your candidate list. Amazon tells you not to repeat them, and it is right. What remains is what the seven fields are for.
- Fill the fields with specific words rather than quoted phrases, with each field covering a different angle: audience, format, occasion, problem, sub-topic, comparison point, setting. There is no benefit to entering the same word twice.
- Price inside the band those result pages showed you, or make the cover earn the difference.
- Two to four weeks in, run an automatic sponsored products campaign at a small daily budget, pull the search term report, and rewrite the fields against what shoppers actually typed. Then diarise it and do it again in six months.
Questions that come up every time
Yes, for one narrow job: learning the vocabulary a topic uses and seeing its rough shape. Not for deciding what goes in seven fields, because they measure what people ask rather than what people buy.
People argue about how much of the description is indexed, and you do not need to settle that to make the right call. The description is the last thing between a shopper and a purchase, so write it to sell. If your terms fit naturally, good. Never sacrifice the first two lines, which are what shows above the fold, to place a keyword.
As many specific, non-duplicated words as the field will accept, and it shows its limit as you type. The costly mistakes are blank fields and fields full of words already in your title, not running one close to full.
The tell
You can spot an imported Google workflow in about ten seconds. The keyword list is full of things people would find interesting to read about, and the listing never gets in front of anyone who arrived with a card in their hand.
The mistake is not using SEO on Amazon. It is assuming Amazon SEO works like Google SEO. Once you understand what each search engine is trying to satisfy, KDP keyword research becomes a very different exercise.
Answer the second question properly and the seven fields more or less fill themselves.
Related: 14 Common SEO Mistakes and Expert Fixes for Higher Rankings
