Amazon Keyword Research: A Working Method

11 August 2026 · 9 min read · AMZNGO

Most keyword research on Amazon produces a spreadsheet nobody uses. The problem is not the tool — it is that the list is built to be comprehensive rather than to be acted on. A keyword you cannot win, or cannot convert once you win it, costs you money in exactly the same way a bad image does. Here is the method we use, and the order it goes in.

Amazon search is not Google search

The distinction that changes everything: Google is trying to answer a question, Amazon is trying to complete a sale. A9 ranks on the probability that a query ends in a purchase — so relevance matters, but conversion history on that exact term matters more. Two consequences follow, and both are counter-intuitive:

• Ranking for a term you convert badly on makes things worse. Impressions without sales tell Amazon the match was wrong, and the term decays. You cannot brute-force your way onto a head term you have no business on.
• You do not need the keyword in the listing more than once. Amazon indexes a term when it appears anywhere in the listing's text fields. Repeating it does nothing except make the copy worse for the human reading it.

Density, keyword stuffing and "putting it in three times to be safe" are Google habits from 2011 that survive on Amazon because nobody tests them.

Step 1 — Start from the product, not the tool

Before opening anything, write down how four different people would search for what you sell. Not synonyms — different mental models:

• By the category name. "resistance bands"
• By the problem. "physical therapy bands for shoulder"
• By the constraint. "resistance bands heavy 150 lb"
• By the occasion. "home gym starter set"

This list is short and it is the only part of the process that requires you to know your buyer. Everything downstream is filtering. Skip it and you get a list optimised for whatever the tool happened to have in its index.

Step 2 — Take the terms Amazon has already given you

The most reliable keyword data for your product is not in any third-party tool. It is in Seller Central, it is free, and it is specific to you:

• Search Query Performance (Brand Analytics, Brand Registry required) — the actual queries that led to impressions, clicks and purchases on your ASINs, with your share of each. This is the closest thing to ground truth that exists.
• Search Terms report from your Sponsored Products campaigns — every literal query a shopper typed that triggered your ad, with spend and orders against it. An auto campaign run for two weeks on a modest budget is the cheapest keyword research available, because you are buying real conversion data rather than an estimate of one.
• Business Reports — session-to-unit conversion per ASIN, which tells you whether a term you already rank for is worth defending.

If you take one thing from this article: run an auto campaign before you write the listing. It costs less than the tool subscription and it answers the question the tool is guessing at.

Step 3 — Mine the competition, carefully

Reverse-ASIN tools (Helium 10, Jungle Scout, Data Dive and the rest) tell you which terms a competing ASIN ranks for. Useful, with one discipline attached: pick competitors that are genuinely comparable. Pulling terms from the category's dominant listing gives you the keyword set of a product with 4,000 reviews and a price you cannot match. You will index for those terms, sit on page nine, convert nothing, and conclude that keywords do not work.

Choose three or four ASINs at roughly your review count, price band and variation depth. The overlap between them is your realistic opportunity set.

Step 4 — Filter down to what you can act on

Now cut the list. Every term survives only if it passes all three:

Test	Cut it if
Relevance	A shopper typing it would be disappointed to land on your page. Returns and negative reviews come from here.
Winnability	Page one is owned by listings with ten times your reviews and a lower price. Park it, do not delete it.
Volume that is real	Estimated volume is under a few hundred a month and it is not highly specific. Long-tail is fine; invented is not.

What survives is usually 20–40 terms, not 400. That is the point. A list you can hold in your head is a list you will actually write to.

Step 5 — Place them, once each

Amazon indexes across several fields, and they are not equal:

• Title — heaviest weight, and the thing a human reads first. Lead with the two or three terms that matter most, in a sentence that survives being read aloud. A title stuffed to 200 characters ranks for more and converts for less.
• Bullets — secondary terms, worked into benefit statements. Never a comma-separated keyword dump; the bullets are the highest-read text on the page.
• Backend search terms — 249 bytes, and the correct home for everything that is indexed but ugly: misspellings, synonyms, Spanish-language variants, alternative unit formats. No commas, no repeats of anything already in the title, no brand names but your own.
• A+ content — indexed for search within Amazon in some contexts, but treat it as conversion surface, not keyword surface. Its job is to answer the objection.
• Image alt / keyword fields — where available, describe the image honestly. Same logic as the web: it is an accessibility field first.

One placement per term. The second appearance buys you nothing.

Step 6 — Re-run it against reality

Keyword research is not a launch task. Thirty days after the listing goes live you have something you did not have before: your own conversion data per query. Pull Search Query Performance again and sort by purchase share.

• Terms converting well above your average — push them up the title, bid harder in ads.
• Terms with impressions and no orders — either the listing is not answering that intent, or the term was never relevant. Both are fixable, and only one of them is a copy problem.
• Terms you never appear for — check indexing before assuming a ranking problem.

Where keywords stop being the problem

This is the part most keyword guides leave out. Ranking gets a shopper to the page. Everything after that is the images, the price, the reviews and the answer to whatever they were worried about. We see listings every month with excellent term coverage and a conversion rate that will not move — and the fix is never another keyword.

If sessions are healthy and units are not, the search work is done and something else is costing you the sale. That is a different diagnosis, and we wrote it up separately: why listings get traffic but don't convert.