Amazon SEO: how Amazon ranks products, and what you can actually influence

Amazon SEO, search engine optimisation for Amazon, is a grand name for a simple idea. A shopper types a search, Amazon decides which products to show and in what order, and you want yours near the top for the searches that lead to sales. Most of what is written about it is either too vague to act on or more precise than anything Amazon has published.

This guide is the hub for our listings and search guides. It explains how ranking works, as far as Amazon has documented it, and how to judge it on profit. Three companion guides go deeper: Amazon keyword research covers finding and choosing search terms, Amazon backend search terms covers the hidden keyword field, and Amazon listing optimisation covers the page itself: title, item highlights, bullets, images and A+.

How Amazon search works: what Amazon has published

Amazon’s UK help for sellers describes the first step plainly: customers search by entering keywords, “which are matched against information such as title and description that you provide for a product"¹. The same page names the title, bullet points, product description, generic keywords (the backend search terms) and the product’s browse node as the inputs you control¹.

Before any of that matters, the product has to be eligible to appear. Amazon lists five common reasons an ASIN can’t be found in search: it is a parent ASIN, it has no browse node, it has no offer, its launch date is in the future, or the listing is suppressed or removed for quality issues³. An out-of-stock product with an expired offer can have a perfect listing and still not appear. Amazon also notes that being indexed only makes an ASIN “eligible (but not guaranteed)” to appear for matching searches².

The second step, putting the matching products in order, is where Amazon says least in its help pages and most in its research. In a 2016 talk at SIGIR, an academic conference on information retrieval, two of its search engineers described how Amazon Search ranked products⁷:

• Ranking models are trained on what customers do. They used “labels based on customers actions, such as purchases, add-to-basket, or clicks.”
• Features include sales and reviews as well as several measures of how closely a product’s text matches the query.
• Behaviour outweighs text. Because hundreds of products can look equally relevant to a query, “behavioral features drive the rankings in Amazon Search to a much larger extent than they do in Web Search,” and “typically, they account for most of the variance reduction” in the models.
• A model usually covers one category in one marketplace, for example Books in Japan, and results from different categories are blended when a shopper searches all departments.

A 2019 paper from Amazon added that matching goes beyond exact words. Pure keyword matching, it said, misses synonyms and broader terms, trips over variants such as “woman” and “women”, and is sensitive to spelling errors, so Amazon trained a model on customer behaviour to retrieve products that are semantically related to a query⁸.

Two cautions. These are research descriptions, the 2016 one now ten years old, not a specification of today’s system. And neither gives weights. What they do establish, from Amazon’s own engineers, is the shape of the thing: text gets you into the set of matching products, and customer behaviour decides the order.

What Amazon doesn’t publish

Amazon does not publish how much each factor counts, how quickly a change is reflected in rank, or how organic and sponsored results interact. Its help pages go out of their way to keep options open: Amazon “reserves the right not to use all supplied keywords”, for reasons including “potential manipulation of search results” and irrelevance², and says the same of the product description and bullet points¹. Which of your search terms count is decided by “a machine learning model” and “may change over time"².

That gap is filled by a lot of confident folklore. We have removed several of these claims from our own older posts because we could not find a primary source.

What Amazon has documented is enough to act on. Be eligible, be relevant in the fields Amazon reads, and convert the traffic you get.

Relevance: the fields Amazon reads

Relevance is the part you write. Each field has its own rules on Amazon.co.uk, and since July 2026 the title has far less room than it used to.

Three practical rules follow from Amazon’s own guidance. Use each important word once, in the most visible field it fits: Amazon tells sellers not to repeat words within the search terms field², and its seller guide warns that keyword stuffing “can create a poor customer experience and hurt your rankings"⁹. Don’t add misspellings or every plural, because Amazon says there is “no need for misspellings” and you should use “singular or plural; no need for both"². And fill in attributes accurately, because a wrong or missing browse node can take a product out of search altogether³.

Our listing optimisation guide covers how to write each field, and the backend search terms guide covers the hidden field byte by byte. Run any title through the listing checker before you publish it.

Performance: conversion and sales velocity

Once a product is relevant, its position depends mostly on whether shoppers choose it. In practice that means four numbers, each of which you can see in your own data.

If you own the brand, the Search Query Performance dashboard in Brand Analytics shows this funnel for up to 1,000 queries: query volume, impressions, clicks, basket adds and purchases, covering organic and Sponsored Products results from the search page¹². Its Query Detail view compares your ASIN with the top 10 ASINs for that search¹². That comparison tells you where to work:

• Seen but not clicked: the problem is what shoppers see in the results. Main image, title, price, rating count or delivery speed.
• Clicked but not bought: the problem is on the page. Images, bullets, A+, reviews, or a price that looked better in the results than it does next to the alternatives.
• Not seen at all: a relevance or eligibility problem. Check the fields above, and check the offer and browse node.

Price and availability belong here too. Amazon’s own seller guide lists adjusting price as one of seven ways to improve search rankings⁹, and a product with no live offer is not searchable at all³. Running out of stock does not just lose today’s sales; it stops the sales history that ranking feeds on, which is how a brand’s own Amazon flywheel stalls.

Advertising sits alongside organic search, not inside it. Sponsored Products can put a product in front of shoppers for a search before it has earned that position organically, and every sale it produces is a sale. Amazon has not published whether, or how much, ad-driven sales lift organic rank, so we treat any organic gain as a bonus to be measured, never as the reason to run an ad. Our PPC strategy guide covers how to structure that spend.

Amazon sales rank: what Best Sellers Rank measures

Best Sellers Rank (BSR) is the number in the “Product information” section of a detail page, and it is the most quoted and most misread metric on Amazon. Amazon.co.uk describes it in one sentence: “The Amazon Best Sellers calculation is based on Amazon sales, and is updated hourly to reflect recent and historical sales of every item sold on Amazon.“¹⁰

Amazon’s seller guide adds the detail¹¹:

• Lower is better, and the rank is relative to other products in the same category.
• Recent sales count more than older ones, though both feed in.
• Page views and customer reviews don’t affect it. A rank “doesn’t necessarily indicate product quality”.
• One product can hold several ranks, one per category it sits in, and ranks differ between Amazon stores.
• The best seller badge goes to products that meet an ordering threshold and rank first in a category with at least 100 products.
• BSR is not search rank. A top BSR “doesn’t always equal a high organic search ranking”, but both relate to sales volume.

So BSR is a useful thermometer: for tracking a competitor’s momentum, sizing a category or spotting seasonality. It is a poor target. It counts units sold, whatever they cost you to sell, so a product can climb the chart while losing money on every order. Amazon’s own guide says a BSR doesn’t drive sales directly¹¹.

Where AI shopping fits

A growing share of searches now go through an assistant rather than a list of results. In the UK that assistant is Rufus. Amazon says it draws on the product catalogue, reviews, community Q&As, detail pages and the web, and that for specific product questions it uses “information found on your product’s detail page"¹³. In the US, Amazon combined Rufus with Alexa+ as Alexa for Shopping on 13 May 2026, letting shoppers ask questions in the main search bar and adding AI-generated overviews at the top of search results and on product pages¹⁴. Amazon says Rufus helped over 300 million customers in 2025¹⁴.

Amazon’s COSMO research shows the direction of travel. It describes a system that mines shopping behaviour to build knowledge graphs of what people buy things for, deployed in Amazon search applications such as search navigation¹⁵. Amazon has not said that COSMO, or anything else, ranks products in Rufus.

For SEO, the practical change is small and important: an assistant answers from the facts on your page, so a listing that states uses, sizes, compatibility and audience in plain text serves both the search box and the assistant. Our Rufus and Alexa for Shopping guide covers how to test and fix what the assistant says about your products.

Don’t buy a rank your margin can’t hold

The usual Amazon SEO question is “how do I rank higher?” The better one is “which rankings are worth what they cost?” Ranking is not free when it is bought with advertising, and a search that brings volume but not margin can leave you busier and poorer.

Take one product at £24.99, which Amazon Ads reports at £20.83 without VAT¹⁶. After cost of goods, Amazon’s fees and fulfilment it keeps £5.52 a unit before advertising, a margin of 26.5% and therefore a break-even ACoS of 26.5%. The figures below are an illustration.

Pushing for page one more than doubles units and would look like a triumph on a sales dashboard and in the Best Sellers Rank. It leaves the business with £2,250 less profit a month. The long-tail plan sells fewer units than the push but makes the most money, because those searches convert at a cost the margin can carry.

The push can still be right, but only as a decision with numbers attached. If the higher organic position holds after the ads stop, the investment may pay back. So set the budget, the target and the end date before you start, measure organic units for the term after you stop, and treat a rank that disappears with the ads as a cost, not an asset. Our product launch guide covers how to plan that kind of investment.

A monthly Amazon SEO routine

This is the order we work in on the accounts we manage, starting with the products that make the most profit, not the most sales.

Amazon SEO done this way is not a separate discipline from running the account. It is listing quality, stock and pricing discipline and advertising judgement, pointed at the searches that make you money. Start with the listing checker for the page itself and the keyword research guide for the searches to aim it at.