Most Amazon keyword research produces a spreadsheet of a thousand terms nobody uses again. A useful list is short, sorted by intent, and built mostly from your own data. Here is the method, free tools included.
Amazon PPC
Most Amazon keyword research produces a spreadsheet of a thousand terms that nobody opens again. A list worth having is short, sorted by intent rather than volume, and built mostly from data you already own. Twenty to forty terms per product will carry almost all of your sales.
This is the method we use on every account, using free sources only. Paid tools have their place and they mostly tell you what your own reports already know, so we start with those.
What is on this page
- Sort by intent, not by volume
- Five free sources, in order
- Your own data beats every tool
- Building the list, step by step
- Why the long tail is worth more than it looks
- Keyword research for a range, not a product
- Where the words actually go
- The other half of the job
- The weekly routine that keeps it current
Sort by intent, not by volume
Volume is the number every tool leads with and the least useful thing to sort on. A term with ten thousand searches a month that never converts costs you money every day. A term with four hundred searches that converts at 20% pays for itself and builds rank on something you can actually own.
So the first job is not gathering, it is sorting. Every candidate term goes into one of four buckets: buying intent, category intent, problem intent, or wrong intent. That single act does more for an account than any tool subscription, and it takes an afternoon.
Five free sources, in order
Your own search term report. Terms that have already produced sales for you. Nothing beats this, and most sellers never mine it properly. The routine is in the search term report guide.
Search Query Performance. If you are brand registered, this shows how many shoppers searched a term, how many saw you, clicked and bought, against the category totals. It is the closest thing to real Amazon search volume anybody can get.
Amazon’s own search bar. Type your product and read what it suggests. Those suggestions come from real searches, and the ones with a qualifier attached are usually the phrases worth having.
Competitor listings. Read the titles and bullets of the products ranking above you. Not to copy them, but to learn which words the category actually uses for the thing you sell.
Your reviews and customer questions. The language people use to describe the problem, which is often nothing like the language you use to describe the product.
The gap worth hunting. Terms where Search Query Performance shows plenty of category searches and very few of your impressions are terms you are not indexed for or not bidding on. That gap list is usually shorter than people expect and it is the highest value output of the whole exercise.
Want us to build the list from your own data?
Send the reports. We will come back with the terms we would target and the ones we would cut, free.
Your own data beats every tool
Third party keyword tools estimate. Your reports do not. Amazon knows exactly what was searched, what was shown and what was bought, and it gives you a version of that in two places nobody else can see: your search term report and, with Brand Registry, Search Query Performance.
Automatic campaigns are the engine here. Running one on modest bids and reading it weekly is the cheapest keyword research available, and it is the only source that finds phrasings you would never have thought to write. That matters more each year as conversational searching widens the tail, which is one of the shifts covered in what good TACOS looks like in 2026.
Building the list, step by step
One. Write down what your product is in three ways: what you call it, what the category calls it, and what a customer describing the problem would call it.
Two. Take each of those into the Amazon search bar and note every suggestion with a qualifier attached, such as a size, material, use case or audience.
Three. Pull your own search term report for the last sixty days and add every term that has produced a sale.
Four. Pull Search Query Performance and add terms where the category gets real volume and you get very few impressions.
Five. Sort everything into the four intent buckets. Delete the wrong intent bucket, or better, keep it as your starting negative list.
Six. Pick the top ten buying intent terms and the top ten category terms. That is your list. The rest lives in automatic campaigns, which is exactly where it belongs.
The list is not the deliverable. The decision about which twenty terms deserve a deliberate bid is the deliverable.
Why the long tail is worth more than it looks
A term with four hundred searches a month feels too small to bother with. Twenty of them are not.
Long tail phrases share three useful properties. They cost less per click, because fewer advertisers bid on them. They convert better, because the shopper has already narrowed what they want. And they are winnable, because a smaller brand can realistically rank first on a four word phrase in a way it cannot on a two word one.
The strategic version of that: start where you can win, use the sales velocity from those wins to build rank, and let the head terms become reachable rather than attacking them on day one with a bid you cannot sustain. That sequence is the whole argument in the first 90 days of a product launch.
Conversational searching has widened that tail further, so the number of viable specific phrases keeps growing. It also means the tail is harder to guess and easier to discover, which is another reason automatic campaigns earn their place.
Keyword research for a range, not a product
Most guides assume one product. Most sellers have a range, and the same research done product by product produces enormous duplication and a lot of internal competition.
Do it once for the range, then split. Shared terms, the ones describing the category, go to your best converting or best reviewed product, because that is the one most likely to turn the traffic into a sale. Differentiating terms, the ones naming a size, colour, capacity or use case, go to the specific variant they describe.
Then check for overlap deliberately. Two of your own products bidding on the same exact term are pushing your own click price up, and the fix is a negative in one of them rather than a bid change in both.
Where the words actually go
Title. Your single most important term, inside the first eighty characters, in a sentence a human would read. That is where mobile truncates, which we covered in your listing is a mobile listing now.
Bullets. Supporting terms in natural language. Bullet one should carry the main reason to buy rather than a specification, because it is often the only bullet a phone user reads.
Backend search terms. Everything that did not fit, without repeating anything already in the title. Amazon indexes a term once. Repeating it wastes the space and does nothing.
A plus content. Text inside images is not readable for indexing, so treat A plus as a conversion tool and let the alt text carry any indexing value. More on that in why your A plus content is not converting.
One warning worth repeating. A title stuffed with terms converts worse, and conversion rate is what actually drives rank. Stuffing costs you twice: nothing gained in indexing, real ground lost in ranking.
The other half of the job
Keyword research is usually framed as finding words to target. Half the value is finding words to block.
Every wrong intent term you spot during research is a negative you can add before it costs you anything. Competitor brand names you do not want to pay for. Materials, sizes or pack counts you do not sell. Research phrases where the shopper is reading rather than buying. Building that list at the same time as your target list is free, and it stops the leak before it starts.
The weekly routine that keeps it current
Full research twice a year per product. In between, twenty minutes a week does more.
Open the search term report for the last sixty days. Look at what your automatic campaigns surfaced. Promote anything converting into its own exact match keyword with a deliberate bid, and add it as a negative in the campaign it came from so the two do not bid against each other. Negate anything past your break even click count with no sale.
That is it. Done weekly, your keyword list stays current without ever sitting down to redo it, and it stays built from what is actually selling rather than what a tool estimated last spring.
Want the whole thing done and kept current?
Twenty minutes on a call and we will show you what your own reports are already telling you.
Send us the reports and we will build the list for you
Free, no card, written answer in three working days, and yours to keep either way.
Frequently asked questions
Far fewer than most lists contain. Twenty to forty terms per product, sorted by intent, will carry almost all of your volume. A thousand row spreadsheet is a research artefact, not a plan, and nobody ever comes back to it.
Amazon’s own search bar suggestions, your Search Query Performance report if you are brand registered, your own search term report, competitor listings, and the questions in your reviews. Between them those five cover most of what paid tools would tell you, and the middle two are data nobody else has.
Phrases. Single words are usually too broad to convert and too expensive to own. Two to four word phrases carry buying intent, cost less per click and are where a smaller brand can realistically win.
The most important in the title, ideally within the first eighty characters. Supporting terms across bullets and description in natural sentences. Everything else in backend search terms, without repeating what is already in the title, since repetition adds nothing.
No, and it costs you twice. Amazon indexes a term once, so repeating it adds nothing, and a title crammed with keywords converts worse, which lowers your rank through the mechanism that actually matters.
Properly, twice a year per product. But the useful work is weekly and small: read the search terms your automatic campaigns surfaced, promote what converts, negate what does not. That routine finds more than any research session.
A brand registered report showing, per search term, how many shoppers searched it, how many saw your product, how many clicked and how many bought, alongside the category totals. It is the closest thing to real Amazon search volume any seller can access.
Send the right reports. Get a straight answer back.
We read the account properly and write back with where the money is going, what it is buying, and the three things we would change first. You keep it either way.
No obligation · no card · a person replies, usually the same working day