Amazon keyword research has one beautiful property: everyone searching is holding a wallet. There are no idle browsers typing "yoga mat with strap for wide shoulders" for fun.

The free method mines Amazon's own suggestion box, because that box is a live feed of what shoppers want to buy in their own words.

Why Amazon's suggestions beat any external tool

Amazon autocomplete is generated by Amazon shoppers, ranked by Amazon's interest in selling things. It encodes demand, phrasing and product features simultaneously.

Type "yoga mat " and the completions are a market research report: thickness numbers, materials, "with strap", "for kids", "non slip". Each suggestion is a feature shoppers filter by.

Mining suggestions at scale

Letter-by-letter manual mining works but crawls. The Amazon tab on this site's tool automates it: seeds in, hundreds of real Amazon suggestions out, free.

This site's keyword tool on the Amazon tab streaming 119 real Amazon suggestions for yoga mat
The Amazon tab on "yoga mat": 119 shopper phrasings, streaming live from Amazon's own suggestions.

Seed with your product type, then again with audience and problem variants: "yoga mat", "yoga mat for", "yoga mat with", "yoga mat no". The prepositions unlock the feature tail.

Reading the patterns

Amazon suggestions sort themselves into four buckets, and each bucket maps to a listing decision:

PatternExampleWhat it tells youWhere it goes
Feature"yoga mat extra thick"Specs shoppers filter byTitle + bullets
Audience"yoga mat for kids"Segments worth targetingBullets, maybe a variant
Problem"yoga mat non slip"The worry to answerTitle if you solve it
Occasion"yoga mat gift set"Seasonal packaging anglesBackend terms

If a feature keeps appearing in suggestions and your product has it, it belongs in the title. Shoppers search their filters.

Where the keywords go

1. Title: main phrase + top features, readable, no stuffing 2. Bullets: features and audiences in shopper phrasing 3. Backend search terms: ~250 bytes, invisible One rule everywhere: each phrase once across the listing. Amazon indexes the set, repetition buys nothing.
Weight flows downward, and the backend field catches synonyms and misspellings you would never print on the page.

The backend field is for the phrasings that would embarrass the listing: misspellings, awkward word orders, regional synonyms. Invisible to shoppers, indexed by Amazon.

Ranking the candidates without paid volume data

Amazon publishes no keyword volumes, so free ranking uses two proxies. Suggestion position: phrases appearing after one letter out-demand phrases needing six.

And result counts: a phrase returning 200 products is a niche, 40,000 is an ocean. Pair high suggestion rank with modest result counts and you have found the pocket.

Google volumes from our free stack add background context for product terms, and paid suites like Helium 10 or Jungle Scout sell Amazon-specific estimates when you outgrow proxies.

The refresh habit

Amazon suggestions move with the market: new features, new worries, seasonal phrasings. Re-mine your seeds quarterly and before Q4, and update bullets when a new pattern earns its place.

Fifteen minutes a quarter keeps the listing speaking this year's shopper language.

The one-line takeaway: Amazon's suggestion box is free market research written by buyers. Mine it in bulk, sort the patterns into features, audiences, problems and occasions, and place each phrase once where it counts.