Amazon keyword research means working out which search terms your product can realistically win and then putting each one in the place where it’ll do the most good: the title, the bullets, the backend field or an exact-match campaign. The best source of data is Amazon’s own, and most sellers don’t start there. They open a tool, type in a seed word and export 3,000 rows.
That gives you a big list. It doesn’t give you a plan. What follows is the method we use on client accounts, in the order we do it.
Where does Amazon keyword data actually come from?
There are two kinds of keyword data, and you need to know which one you’re looking at. First-party data comes from Amazon: Brand Analytics, the Search Query Performance report and your advertising reports. It’s measured, but it only covers your own brand and your own ads. Third-party data comes from tools like Helium 10, Jungle Scout, SellerSprite and Data Dive. It covers the whole market, but it’s estimated.
Neither is complete on its own, which is why the order you use them in matters.
| Source | What it gives you | Measured or estimated | Covers competitors? |
|---|---|---|---|
| Search Query Performance (Brand Analytics) | Impressions, clicks, basket adds and purchases per query, total vs your brand | Measured | No — your brand only |
| Top Search Terms (Brand Analytics) | Search Frequency Rank, top 3 clicked ASINs, their click and conversion share | Measured (relative rank, not volume) | Yes — top 3 per term |
| Sponsored Products search term report | Real customer search terms that triggered your ads, with spend and sales | Measured | No |
| Search Term Impression Share report | Your impression rank and share on search terms you advertised on | Measured | Indirectly |
| Reverse ASIN tools (Cerebro etc.) | Keywords a competitor ASIN appears for, estimated volume, sampled rank | Estimated | Yes |
| Keyword tools (Magnet, Keyword Scout etc.) | Related terms and estimated volume from a seed | Estimated | N/A |
Brand Analytics needs a Professional selling account and Brand Representative access on a brand enrolled in Brand Registry. It’s free, it sits under Brands → Brand Analytics, and it holds roughly three years of history, refreshed within about three days.
If you aren’t Brand Registered yet, the method still works. You just lean harder on your search term report and the tools.
Step 1: What are shoppers already using to find you?
Start with Search Query Performance, not a keyword tool. SQP shows the top queries that led customers to your products, and how you did at each stage: impressions, clicks, basket adds, purchases, with your brand’s share of each. That’s the list of searches that already pay you. Everything else is a hypothesis.
Pull the last 13 weeks at brand level and then at ASIN level. Sort by purchases, not impressions. Then tag each query with where you’re losing people:
- High impression share, low click share. Shoppers see you and pick someone else. That’s a main image, price or title problem, not a keyword problem.
- Good click share, low basket-add share. They land on the page and leave. Check the listing against the query — does the page actually answer what they searched for?
- Good basket adds, low purchase share. Usually price, delivery promise or reviews at the last step.
- Low impression share on a query that converts well when you do show. This is your keyword opportunity. You’re relevant and you convert, you just aren’t visible enough.
That last group is the one the rest of this method builds on. Write it down before you look at a single competitor, because once you’ve seen their keyword list you’ll anchor on it.
Step 2: How do you see what competitors rank for?
Run a reverse ASIN lookup on the three to five products currently taking the positions you want. Not the category leader by default — the products that show up above you on your best-converting queries from Step 1. The Top Search Terms report tells you who they are: for any search term, it lists the three most-clicked ASINs with their click share and conversion share.
Then put your ASIN and theirs into a multi-ASIN comparison (Cerebro does up to ten). The useful output isn’t what they rank for. It’s what several of them rank for and you don’t.
Two things to keep in mind, both covered in more detail in our reverse ASIN lookup guide:
- The volume figures are modelled. Helium 10 describes its search volume as an estimate projected from recent weekly velocity. Useful for comparing terms against each other. Not a count.
- The rank figures are sampled. “Organic rank 8” is where the tool last detected the product, on a results page that changes by shopper, location and time of day.
So treat the tool output as a shortlist, not a verdict.
Step 3: How do you check whether a keyword is worth chasing?
Score each candidate on four things: relevance, demand, competition and conversion evidence. Volume on its own is the metric most people sort by, and it’s the one that misleads most often. A 40,000-search term that your product only half-matches will cost you ad money and give you nothing back.
Here’s the scoring grid we use. It’s a working heuristic, not an Amazon rule.
| Factor | Question | Where to check | Score 0–3 |
|---|---|---|---|
| Relevance | Would a shopper typing this be happy to land on our product? | Your judgement + the page-one results | 0 = no, 3 = exact match |
| Demand | Is the search big enough to matter? | Search Frequency Rank first, tool volume second | 0 = negligible, 3 = top of category |
| Competition | Can we get on page one? | Title density, review counts on page one, your impression rank | 0 = locked, 3 = open |
| Conversion evidence | Has this term (or a close variant) converted for us? | SQP purchases, search term report orders | 0 = no data, 3 = proven |
Anything scoring 9 or more goes into the plan. 6–8 goes into a discovery campaign to earn its evidence. Under 6 gets dropped for now.
Check demand against Amazon, not the tool. Top Search Terms gives a Search Frequency Rank for each term. It’s a rank, not a volume — rank 1 is the most searched term on Amazon that period, and Amazon doesn’t say how many searches that is. But when a tool claims a term has 25,000 searches a month and its Search Frequency Rank sits in the hundreds of thousands, trust Amazon.
Title density is the cheapest competition check. It’s the number of page-one products with the term in their title. Decent demand with low title density means page one isn’t really optimised for that phrase. Those are the terms you can win without outspending anyone.
An illustrative example
To show the grid working, here’s a made-up product — a 1-litre insulated steel water bottle — with made-up figures. It isn’t client data.
| Keyword | Relevance | Demand | Competition | Conversion evidence | Total | Decision |
|---|---|---|---|---|---|---|
| water bottle | 1 | 3 | 0 | 1 | 5 | Drop: too broad, page one is locked |
| insulated water bottle | 3 | 3 | 1 | 3 | 10 | Title + exact-match campaign |
| 1 litre water bottle metal | 3 | 2 | 2 | 2 | 9 | Title or Item Highlights + exact |
| gym water bottle 1l | 2 | 2 | 3 | 1 | 8 | Discovery campaign first |
| leak proof bottle for school | 1 | 2 | 2 | 0 | 5 | Drop: wrong buyer |
The highest-volume term lost. That’s normal.
Step 4: Where does each keyword go on the listing?
Put each keyword where Amazon and the shopper will both read it, and never pay for the same space twice. Since 27 July 2026, titles in every category except Media are capped at 75 characters including spaces, with up to 125 characters of extra detail moving into a separate Item Highlights field. That changes how you place keywords more than any tool update this year. Our 75-character title post covers the rule itself.
Our placement order, as of September 2026:
- Title (75 characters). The one or two highest-scoring terms, written so a person can read it. The rule that no word appears more than twice (excluding articles, prepositions and conjunctions) still applies.
- Item Highlights (up to 125 characters). The next tier: size, material, use case, compatibility.
- Bullet points. Supporting terms, in sentences that answer the questions from the SQP funnel gaps in Step 1.
- Backend search terms (249 bytes). Everything relevant that didn’t fit naturally anywhere visible: synonyms, alternative spellings, other-language variants, use cases.
- A+ content. Written for conversion. Don’t rely on it for indexing.
Backend search terms: the rules that catch people out.
- The limit is 249 bytes, not characters. Standard letters and numbers are one byte each; accented letters and symbols can take two to four. A field that looks short can still be over.
- Going over is widely reported to stop the field indexing, not just the overflow. Check the byte count before you save.
- Separate words with spaces. Commas and other punctuation use bytes and add nothing.
- Don’t repeat words already in the title or bullets, and don’t add plurals or capitalised versions — Amazon handles those itself.
- Leave out brand names (yours and anyone else’s), ASINs, and claims like “best” or “new”. Competitor brand names in particular can get the listing flagged.
Then check it indexed. Search Amazon for your ASIN plus the keyword (for example B0XXXXXXXX insulated water bottle). If your product comes up, the term is indexed. If it doesn’t, the term isn’t counting, whatever the backend field says. Leave 24–48 hours after an edit before you test.
A quick note on Rufus and COSMO, because you’ll read a lot about them. Amazon’s search now layers an intent-understanding system (COSMO) and a conversational assistant (Rufus) on top of classic keyword matching. Amazon hasn’t published how either one weighs listing text. The practical takeaway we can defend: listings that plainly state who the product is for, what problem it solves and how it’s used give both systems more to work with than listings stuffed with keyword variants. Write for the question, not just the phrase.
Step 5: How do you use PPC to test and defend keywords?
Use campaigns to buy data on the keywords you aren’t sure of, and to hold position on the ones you’ve proven. Organic rank on Amazon follows sales on a keyword. PPC is the fastest way to generate those sales while the listing changes bed in.
The routing we use:
- Score 9+ and already ranking organically on page one: exact-match campaign at a defensive bid. You’re protecting rank, not buying it.
- Score 9+ and not yet ranking: exact match, single-keyword ad group, bid for top of search. This is where a SKAG structure earns its keep — one term, one bid, one clear read on cost.
- Score 6–8: a phrase or broad discovery campaign with a tight negative list. Harvest the converting search terms into exact after 14–30 days.
Two reports tell you whether it’s working:
The Sponsored Products search term report shows the actual customer searches that triggered your ads. It’s the most honest keyword research data you own, because every row has real spend and real orders against it. It also shows you where broad match wanders off.
The Search Term Impression Share report shows your impression rank and impression share for search terms you advertised on, over up to the last 30 days. If a term converts well but your impression share is low, you’re under-bidding a proven keyword. If your share is high and the term still doesn’t convert, more bid won’t help.
Once a term holds organic page-one rank for several weeks, test lowering its bid. If organic sales on the term hold, keep the saving.
How often should you redo Amazon keyword research?
Rerun Step 1 monthly and the full five steps quarterly, plus any time sessions drop without an obvious cause. Search behaviour shifts with seasons, new competitors take page-one spots, and Amazon changes listing rules — the title limit this year is the obvious example.
A monthly check takes under an hour:
- Pull SQP for the last four weeks and compare impression share on your top 20 queries against the previous month.
- Check the top 3 clicked ASINs in Top Search Terms for your five most valuable queries. New names are new competitors.
- Re-run the indexing check on your title and backend terms.
- Harvest converting search terms from discovery campaigns into exact.
If impression share has dropped on queries where your conversion rate hasn’t, you’ve lost visibility, not appeal. That pattern has its own diagnostic, which we cover in why sales drop when traffic looks flat.
The common mistakes
Sorting by volume. The biggest terms are usually the least relevant and most expensive. Relevance and conversion evidence decide whether a term pays.
Using one tool’s numbers as fact. Tools disagree with each other by wide margins on the same term. Use them for relative comparison and check the important ones against Search Frequency Rank.
Keyword stuffing the title. It breaks the two-repeats rule, it reads badly to shoppers, and under the 75-character limit there’s no room for it anyway.
Doing research once, at launch. The keyword set that launched the product isn’t the one that will grow it. Your own SQP and search term data after 90 days is worth more than anything you had before launch.
Researching without placing. A 3,000-row export with no decision about where each term goes is a spreadsheet, not a strategy.
FAQ
What is Amazon keyword research?
Finding the search terms shoppers use for your type of product, working out which ones you can realistically win, and placing each one in the title, bullets, backend search terms or a PPC campaign. The aim is visibility on searches that convert, not the largest possible keyword list.
What is the best free Amazon keyword research tool?
Amazon’s own Brand Analytics, if you’re Brand Registered. Search Query Performance gives measured impressions, clicks, basket adds and purchases for your brand’s queries, and Top Search Terms gives Search Frequency Rank plus the top three clicked products for any term. Your Sponsored Products search term report is free for any advertiser.
Does Amazon show keyword search volume?
Only for your own brand’s top queries. Search Query Performance in Brand Analytics shows search query volume, the number of times customers searched for a query, for the top queries that led customers to your products, by week, month or quarter. For any other term, Brand Analytics gives only Search Frequency Rank, a relative ranking where 1 is the most searched term in the period, without the number of searches behind it. Monthly volume figures in third-party tools are estimates.
How many characters can Amazon backend search terms have?
The limit is 249 bytes, not characters. Plain letters and numbers are one byte each, but accented letters and symbols can use two to four bytes. Going over is widely reported to stop the whole field from indexing, so check the byte count before saving.
How do I know if my Amazon keywords are indexed?
Search Amazon for your ASIN followed by the keyword. If your product appears, the term is indexed for that listing. If it doesn’t, it isn’t counting. Allow 24–48 hours after a listing edit before checking.
How often should I update my Amazon keywords?
Review your Search Query Performance data monthly and redo the full research quarterly. Also rerun it whenever sessions fall without a clear reason, when a new competitor takes a top-three click position, or when Amazon changes listing rules.
Sami Sultan runs Ecom Enable, an Amazon growth partner for DTC brands. Report definitions and listing limits checked 22 September 2026. If you’d like this method run on your own account, it’s part of our free Amazon audit.
Sources: ZonGuru on Brand Analytics reports, eligibility and data retention; Keywords.am on Top Search Terms and Search Frequency Rank; Amazon Ads: Search term impression share report for Sponsored Products; Intentwise on the SP Search Term Impression Share report; Amazon Ads: Search term report for Sponsored Products; SellerSprite on backend search term byte limits; Perpetua on search term guidance; Helium 10 on how search volume is estimated. Title limit: Ecom Enable’s 75-character title post, sourced to Seller Central.