Amazon SEO is the work you do on product listings so they show up higher when someone searches on Amazon. On Google the page that best answers a question wins. On Amazon the product that people actually buy for that search wins: how relevant the text is and how much it really sells count together.
In the UK the stakes are high. Around 100,000 British small and medium-sized businesses sell on Amazon, and in 2024 they made over £3.8 billion in export sales (Amazon UK, checked on 10 October 2026). Yet the guides on page one of Google UK repeat advice and US statistics without sources. Here you'll find the current official rules, what Amazon has actually published about how its search works, and a comparison we haven't seen elsewhere: the same products searched on Google and on Amazon in the UK.
Amazon SEO in the UK: the numbers to start from
According to the Office for National Statistics, online sales made up 27.5% of all retail sales in Great Britain in 2025, and 27.3% in August 2026 (release of 18 September 2026). More than a quarter of every pound spent in British shops is spent online. A large share of that goes through a marketplace, and Amazon is the biggest.
Amazon's own page on UK small businesses adds a few details that matter if you sell:
- around 100,000 UK SMEs sell on Amazon;
- over £3.8 billion in export sales in 2024;
- around 60% of Amazon's sales come from small businesses, which sold more than 1 billion products in 2024 (a figure Amazon gives without saying whether it covers the UK alone);
- many of those UK sellers are based outside London; Amazon doesn't publish a breakdown by region.
One figure many guides take for granted is missing: how many product searches start on Amazon rather than on Google. The figure quoted most often is American. In the PowerReviews Power of Reviews 2023 survey of 8,153 US consumers, 50% said they start a product search on Amazon.com and 31.5% on Google. For the UK we found no equivalent public survey, so that 50% should be read as a US figure and nothing more.
How search works on Amazon
In its official guide for sellers, published in September 2025, Amazon describes its search as a system in five parts: the search bar, filters, the results page, sales rank and sponsored products. The listing text gets you into the results. Price, availability, reviews and sales decide where you appear.
There's some confusion about the algorithm's name. A9 was an Amazon subsidiary, founded in 2003, that built product search technology; in 2019 its site was closed and the work moved into Amazon Search. The habit of calling the internal engine A9 stuck. "A10" is a label used by consultants and software vendors: it doesn't appear in any of the official sources we checked for this article. What Amazon has published about how search works matters more than the name.
What Amazon Science research says
In 2019 a group of Amazon researchers, including Priyanka Nigam and Vijai Mohan, presented the paper Semantic Product Search at the KDD conference. Their starting point is that matching a query to a listing word by word isn't enough: it doesn't understand synonyms and antonyms, it breaks on singular and plural and it struggles with typos. The semantic model they proposed improved Recall@100 by at least 4.7% and mean average precision (MAP) by 14.5% over the baseline systems, results later confirmed with A/B tests in production.
In 2024 Amazon published COSMO (SIGMOD conference), a system that builds a common-sense knowledge graph from user behaviour: it links a product to the uses, occasions and needs it's bought for. It covers 18 major categories and was deployed in search navigation, with improvements measured in A/B tests.
The practical consequence is simple. Repeating the same keyword in every possible variant helps less and less, because the engine works out the variants by itself. What helps is a listing that says what the product is for and who it's for. Google made the same shift years ago, as we explain in our guide to how search engines work.
Indexing and ranking are two different things
In a guide published on Search Engine Land in December 2024, Robyn Johnson shows a listing indexed for 1,923 keyword phrases according to Helium 10, but on page one for only 80. Being findable for a word doesn't mean being visible: to climb, the listing has to sell on that specific search. That's why new listings nearly always start from long-tail keywords, where there's less competition, as we explain in our article on long-tail and short-tail keywords.
Same keyword, two engines: Google and Amazon in the UK
Many sellers choose the words for their listing by looking at Google volumes. We compared 13 common product searches in the UK: the average monthly volume on Google, from Google Ads data, and the estimate of searches on Amazon.co.uk calculated by DataForSEO Labs.
| Search | Google UK (searches/month) | Amazon.co.uk (estimated searches/month) | Google/Amazon ratio |
|---|---|---|---|
| air fryer | 201,000 | 25,170 | 8.0 |
| coffee machine | 165,000 | 8,452 | 19.5 |
| mattress | 135,000 | 73,341 | 1.8 |
| treadmill | 74,000 | 8,265 | 9.0 |
| backpack | 74,000 | 10,172 | 7.3 |
| running shoes | 74,000 | 18,737 | 3.9 |
| smart watch | 60,500 | 10,584 | 5.7 |
| power bank | 49,500 | 8,115 | 6.1 |
| wireless headphones | 40,500 | 11,556 | 3.5 |
| robot vacuum | 40,500 | 1,084 | 37.4 |
| dog food | 33,100 | 16,792 | 2.0 |
| cat litter | 18,100 | 4,562 | 4.0 |
| wood pellets | 5,400 | 1,234 | 4.4 |
A note on method first. Amazon doesn't publish search volumes for its marketplace: the figures in the third column are estimates from an outside provider, not Amazon data. Read them as orders of magnitude and for relative comparisons, not as exact numbers.
What interests us is the ratio, and the ratio isn't constant. For "mattress" and "dog food" demand on Amazon is about half of Google's, close by the standards of this table: these are products British shoppers clearly look for directly on the marketplace. For "robot vacuum" Google gets over 37 times the estimated Amazon searches, and for "coffee machine" almost 20 times: these are purchases people research with reviews, comparisons and brand sites before they reach a marketplace, and on Amazon they often search by brand or model instead. For sellers, this means a strong word on Google can be weak on Amazon and the other way round. The right data sit inside Amazon: if your brand is enrolled in Brand Registry you get Brand Analytics, which shows how customers search and buy. For the Google side, our selection criteria are in our article on keyword research tools.
The listing, field by field
Title: the rules in force since 21 January 2025
In January 2025 Amazon published its new title requirements on the Seller Central Europe forums, applying "across our stores" to most categories:
- a maximum of 200 characters, including spaces;
- the characters ! $ ? _ { } ^ ¬ ¦ are banned, unless they're part of the brand name;
- the same word can't appear more than twice, except prepositions, articles and conjunctions.
Brand owners who don't comply get a suggested correction and have 14 days to act on it. After that, Amazon updates the title itself. The seller guide recommends staying under 60 characters, with product type, brand and descriptive details. The repetition rule officially ends the era of titles built as lists of keywords, a habit that on Google is called keyword stuffing and that hasn't paid off there for years either.
Bullet points
Bullet points are the part customers actually read before they buy. Amazon suggests opening each point with a one- or two-word description and explaining it in 100 characters or fewer. Write them with real questions in mind: size, compatibility, what's in the box, what the product is designed for. That's also the information COSMO and Amazon's AI assistant use to match a product to a request.
Description and A+ content
With a registered brand, you can replace the text description with A+ content, made of modules combining images and text. On its Brand Registry page Amazon claims, from internal data, a sales increase of up to 8% for basic A+ and up to 20% for Premium A+. On the same page Amazon says that optimising listings with tests in the Manage Your Experiments tool can lift sales by up to 25%. These are the vendor's figures, and they're maximums, not averages.
Hidden keywords (search terms)
In the back end of the listing there's a field for keywords the customer doesn't see. Amazon asks for them in lower case, separated by spaces, without articles or prepositions. The length limit, expressed in bytes, is set out in the Seller Central help pages: check it there before you upload. Use the field for synonyms, regional terms ("trainers" and "sneakers", "hoover" and "vacuum"), different formats, common misspellings and words in the languages of the other markets you sell in.
Images
Amazon recommends that the product fills at least 85% of the frame in the main image and that image alt text is a simple sentence with one or two keywords. The same principles apply to a website, as we explain in our guides to image optimisation and alt text. In sponsored products video carries weight: according to Amazon Ads, Sponsored Products campaigns with video had a 67% higher click-through rate and a 9% higher conversion rate (internal data, United States, June 2026).
Reviews, price and sales: the factors you can't write
The heaviest part of ranking isn't in the text. In the official guide price is one of the seven actions listed for climbing the results, and sales rank is one of the five parts of search. Reviews count for both the engine and the customer: in the PowerReviews survey 45% of US consumers say they won't buy a product with no reviews, and 98% are more likely to read them when it's a product they've never bought before.
This is where Amazon SEO stops being a copywriting job. A perfect listing with an uncompetitive price, few reviews and stock that runs out every fortnight won't climb. Conversion rate is the metric to watch every week, keyword by keyword when the data allow. On reviews the usual rule applies: ask for them in the ways Amazon permits, never buy them. In the UK buying fake reviews is also against the law: the Digital Markets, Competition and Consumers Act 2024 lists fake reviews among the practices banned in all circumstances, a rule in force since 6 April 2025 (Schedule 20, paragraph 13). We cover the topic in our article on online reviews.
Rufus: the AI assistant that reads listings for the customer
Rufus is Amazon's generative AI shopping assistant. It launched in the United States on 1 February 2024 and in the UK on 4 September 2024, in beta for a subset of customers in the Amazon mobile app, with a wider rollout in the following weeks (Retail Insight Network; Amazon's announcement to sellers). According to Amazon, it answers by drawing on the product catalogue, customer reviews, community Q&As and information from across the web.
The numbers show it isn't a side experiment. In its fourth-quarter 2025 results Amazon says Rufus was used by more than 300 million customers and contributed almost $12 billion in annualised incremental sales. In the United States, since 13 May 2026, it has been called Alexa for Shopping; the announcement of the new name covered US customers, and Amazon hasn't said when or whether it will reach the UK.
For whoever writes the listings, the kind of question to answer changes. A customer who asks the assistant "what do I need for winter hiking in the Lake District" isn't typing a keyword. The winning listing is the one that says clearly what use, what weather and what kind of person the product is designed for, and whose reviews match what it promises. It's the same logic as Generative Engine Optimisation applied to a marketplace, and it's part of what we call search everywhere optimisation.
Organic and advertising: the same results page
On Amazon, sponsored results are mixed in with organic ones, and advertising has become a huge business: in the fourth quarter of 2025 Amazon's worldwide advertising services revenue was $21.3 billion, up 23% on the year before. For a new listing, Sponsored Products ads help generate the first sales on a keyword, and those sales help the organic position. The reasoning is similar to what we do on Google with strategies that make SEO and Google Ads work together: the paid campaign buys the data and the time that organic doesn't have yet.
Amazon competes with you on Google too
There's a part of Amazon SEO people rarely talk about: Amazon ranks on Google for a huge number of product searches and takes traffic from manufacturers' own sites. In the UK, amazon.co.uk is in Google's top 10 for about 2.1 million keywords (DataForSEO Labs, 10 October 2026). It isn't unbeatable. In the Macropix case, an Italian manufacturer of LED walls, in July 2026 the client's site had 25.55% share of voice on the "ledwall" keyword group on Google Italy, against 18.08% for amazon.it (Semrush data). A page written by the people who make the product beats a marketplace listing when the search calls for explanations and not just a price. If you sell both on Amazon and on your own site, ecommerce SEO and category pages are where you play this game.
The listing isn't yours
The 2025 title rule contains a detail almost nobody comments on: after 14 days Amazon can rewrite your title itself. The same goes for everything else. The reviews, the sales history and the position you earn stay inside Amazon, and one change of policy or interface, like the switch from Rufus to Alexa for Shopping in the United States, is enough to change the ranking rules. That's why businesses that do well on Amazon build a channel of their own alongside it: a website that ranks on Google with the same words and the same product knowledge. If you want to work out where to start with your own site, that's the work we do in our SEO consultancy.
Frequently asked questions about Amazon SEO
Amazon describes its search as a system in five parts: search bar, filters, results page, sales rank and sponsored products. The listing text decides which searches a product appears for; price, reviews, availability and sales on that search decide its position. Research published by Amazon Science, such as Semantic Product Search (2019) and COSMO (2024), shows the engine interprets synonyms, variants and intended uses, not just exact words.
Since 21 January 2025, for most categories, a title can have at most 200 characters including spaces, can't contain the characters ! $ ? _ { } ^ ¬ ¦ (unless they're part of the brand) and can't repeat the same word more than twice, excluding articles, prepositions and conjunctions. Amazon still recommends staying under 60 characters.
Google rewards the page that best answers a question and uses signals such as links and site authority. Amazon rewards the product that sells most on a specific search: conversion, price, reviews and availability count as much as the text. Demand differs too: in our comparison of 13 UK searches the ratio between Google volumes and Amazon.co.uk estimates ranges from about 2 to over 37.
Yes. Amazon launched Rufus in the UK on 4 September 2024, in beta in the mobile app, with a wider rollout in the following weeks. It answers using Amazon's catalogue, customer reviews, community Q&As and information from the web. In the United States the assistant has been called Alexa for Shopping since 13 May 2026.
It isn't compulsory, and it requires a brand enrolled in Brand Registry. According to the internal data Amazon cites, basic A+ can increase sales by up to 8% and Premium A+ by up to 20%. It matters mainly for conversion, which in turn affects position.
As a starting point, yes, but with care. In the data we collected, searches such as "mattress" and "dog food" have relatively close demand on the two engines, while for "robot vacuum" Google gets over 37 times the estimated searches on Amazon.co.uk. If you have a registered brand, you'll find Amazon's internal search data in Brand Analytics.
