Key takeaways
Amazon ranks products to maximize the likelihood of a sale, weighing keyword relevance against purchase behavior. The original system was named A9 after Amazon's now-dissolved A9.com search subsidiary. Sellers call today's more behavior-driven system "A10," but Amazon has never confirmed that name or a version change. Amazon's published AI layer - the COSMO knowledge graph and the Rufus shopping assistant - plus Sponsored ad auctions now also shape what shoppers see.
- A9 is the only ranking system Amazon has formally named, after its A9.com subsidiary (founded 2003, domain retired 2019); "A10" is seller and agency terminology, not an Amazon-confirmed version.
- Amazon officially states that search terms, price, availability, selection, and sales history help determine search placement, and that better-selling products tend to rank higher.
- COSMO (an Amazon Science common-sense knowledge graph) and Rufus (Amazon's generative AI shopping assistant) are confirmed AI systems that influence discovery and search navigation.
- Sponsored Products, Sponsored Brands, and Sponsored Display run cost-per-click auctions; the sales and reviews they generate can indirectly strengthen a listing's organic standing.
- Treat any specific A10 "weighting" as practitioner consensus from observed behavior, not as Amazon's published mechanics.
What "Amazon's ranking algorithm" actually means
In plain terms
Amazon's ranking algorithm is the set of systems Amazon uses to order products in search results so the listings a shopper is most likely to buy appear first. Amazon does not publish a ranking rulebook, so everything beyond a handful of high-level statements is inferred from observed behavior. Throughout this hub we mark each claim as Amazon-confirmed or industry consensus.
In practice, four overlapping systems shape what a shopper sees: the A9 foundation, the behavior-heavy system practitioners call A10, the COSMO and Rufus AI layer, and the Sponsored ad auctions. They are not four separate ranked lists; they interact, and ad performance can feed organic rank.
Amazon does not publish how it ranks products, and it has never released a ranking rulebook the way it documents advertising or listing policies. Everything beyond a handful of high-level statements is inferred by sellers, agencies, and researchers from observed behavior. Throughout this guide we mark each claim as Amazon-confirmed or industry consensus so you never mistake a community theory for official mechanics.
The one thing Amazon is explicit about is its goal: surface the products a shopper is most likely to buy. Amazon's own seller guidance states that search terms, price, availability, selection, and sales history help determine where a product appears, and that "better-selling products tend to be towards the beginning of the results list." That single sentence is the foundation everything else builds on. If you want the consolidated working list of every factor across all four systems, jump to our Amazon ranking factors checklist.
Amazon ranking at a glance
- Goal
- Surface products most likely to sell
- Only named system
- A9, after the A9.com subsidiary
- A9.com founded
- 2003; domain retired 2019
- "A10"
- Seller terminology, not Amazon-confirmed
- AI layer
- COSMO knowledge graph and Rufus assistant
- Ad formats
- Sponsored Products, Brands, Display
- Confirmed factors
- Search terms, price, availability, selection, sales history
- Bedrock
- AWS dev service, not the ranking algorithm
The A9 era: the only officially named system
A9 is the one ranking system tied to a real, named Amazon entity. A9.com was an Amazon subsidiary founded in 2003 and launched in 2004 in Palo Alto, California, built to develop search and search-advertising technology. Its name is a numeronym for "algorithm" - the letter A plus the nine remaining letters. A9 powered product search across Amazon's retail sites for years.
In 2019 Amazon took down the A9.com site and redirected the domain to its homepage, and the function was absorbed into Amazon's internal Search organization. That is why "A9" survives as shorthand for Amazon's product-search ranking even though the standalone entity no longer exists.
Conceptually, the A9 era is usually described around two pillars:
- Relevance - how well a listing's text (title, bullets, description, and backend search terms) matches the shopper's query.
- Performance - signals that predict a sale, above all conversion rate and sales history.
Want the full history and the relevance-versus-performance breakdown? See the spoke guide: The A9 Algorithm Explained.
A10: the behavioral system sellers describe (not Amazon-confirmed)
"A10" is industry and seller terminology - Amazon has not confirmed an algorithm-version change or the name itself. What practitioners label A10 is the same Amazon search system observed to lean more heavily on shopper behavior and signals outside a seller's direct control than the earlier A9 description did.
The consensus factors agencies attribute to the current system include:
- Conversion rate and sales velocity - the speed and volume at which a listing actually sells.
- Keyword relevance - how well listing copy and backend terms map to real queries.
- Customer satisfaction - reviews, ratings, return rate, and seller account health.
- External traffic - off-Amazon clicks (from Google, social, email) that practitioners report can boost standing, though Amazon does not document this.
The honest framing: A10 is a useful label for a set of observed behaviors, not a published Amazon product. Treat any precise "weighting" as a hypothesis. The deep dive lives in the spoke guide: The A10 Algorithm Explained.
The AI layer: COSMO and Rufus
Two AI systems are confirmed by Amazon and increasingly shape discovery. COSMO is a system documented by Amazon Science that mines "user-centric commonsense knowledge" from shopper behavior and large language models to build industry-scale knowledge graphs. It connects products to the situations and intents behind a search (for example, that a query implies an occasion or use case), and Amazon reports it is deployed in Amazon search applications such as search navigation, generating knowledge across 18 product categories with A/B-tested improvements.
Rufus is Amazon's generative AI shopping assistant, launched in beta in early 2024. Amazon states it is trained on Amazon's product catalog, customer reviews, community Q&As, and information from across the web, and it answers conversational questions, compares options, and recommends products inside the shopping experience. (In May 2026 Amazon announced it was bringing Rufus and Alexa+ together into "Alexa for Shopping," retiring the standalone Rufus brand while folding its product-knowledge capabilities into the unified assistant.) For sellers this means listings increasingly need to satisfy a model answering a question, not just match a keyword string.
Note one common point of confusion: Amazon Bedrock is AWS's managed foundation-model service for developers - it is not Amazon's retail ranking algorithm. Rufus is a retail-facing assistant, not Bedrock itself. Read the spoke: Rufus and Amazon's AI Search Layer.
The ad auctions: Sponsored Products, Brands, and Display
Paid placements are a separate system from organic ranking, but they sit in the same results and feed back into it. Amazon's Sponsored Products are cost-per-click (CPC) ads that promote individual listings; advertisers set a bid (the maximum they will pay per click) and Amazon's system weighs that bid against keyword relevance and the shopper's query. Ads can appear at the top of, alongside, or within shopping results and on product detail pages. Sponsored Brands (banner and brand placements) and Sponsored Display (behavior-targeted ads on and off Amazon) round out the suite.
The crucial interaction: ads buy visibility, and visibility drives sales and reviews, which are organic ranking inputs. A well-run launch campaign can generate the early conversion velocity that helps a new listing earn organic position - this is why ads and SEO are managed together on Amazon, not in isolation. The relationship is widely reported by practitioners; Amazon does not publish a formula tying ad spend to organic rank.
Full breakdown in the spoke: Amazon Sponsored Ads and the Auction.
How the systems interact
These layers are not independent. A practical model of the flow:
- Relevance gates entry. If your title, bullets, and backend terms do not match the query, neither organic ranking nor ads can surface you for it.
- Behavior decides order. Among relevant listings, conversion rate and sales history push the best sellers up - the through-line from A9 to A10.
- Ads accelerate behavior. Sponsored placements manufacture impressions and clicks; the resulting sales and reviews become organic signals.
- External traffic and AI compound it. Off-Amazon traffic and strong, answer-ready content help both conversion and AI-surface visibility (COSMO navigation, Rufus answers).
The strategic takeaway: optimize the listing for relevance and conversion first, then use ads to seed early velocity, then keep reviews and content strong so the AI layer can recommend you. No single lever wins alone.
How Amazon's ranking systems evolved: a timeline
Amazon's ranking story runs from the A9.com search subsidiary in 2003 to the generative AI shopping layer of the mid-2020s.
-
2003
A9.com founded
Amazon establishes A9.com as a subsidiary in Palo Alto to build search and search-advertising technology; the name is a numeronym for "algorithm."
-
2004
A9 search launches
A9 goes live and over time powers product search across Amazon's retail sites - the origin of the "A9" shorthand sellers still use.
-
2019
A9.com domain retired
Amazon takes down the A9.com site and redirects the domain to its homepage; the search function is absorbed into Amazon's internal Search organization.
-
2024
COSMO documented
Amazon Science publishes COSMO at SIGMOD 2024, a common-sense knowledge-graph system mined with LLMs and deployed in Amazon search navigation across 18 product categories.
-
2024
Rufus launches in beta
Amazon introduces Rufus, a generative AI shopping assistant trained on its catalog, reviews, community Q&As, and web information, rolling out in the mobile app.
-
2026
Rufus and Alexa+ merged into Alexa for Shopping
In May 2026 Amazon announces it is bringing Rufus and Alexa+ together into "Alexa for Shopping," retiring the standalone Rufus brand and consolidating its conversational shopping AI.
Amazon ranking signals: confirmed vs. consensus
Amazon's ranking systems weigh many inputs, but only some are stated by Amazon. The table below separates what Amazon confirms from what the industry infers, so you do not treat a community theory as published mechanics.
| Signal | What it covers and how confirmed |
|---|---|
| Keyword relevance | Amazon-confirmed. Search terms in the title, bullets, description, and backend fields are matched against shopper queries; better-chosen terms increase visibility. |
| Sales history and velocity | Amazon-confirmed that sales history helps determine placement and better-selling products rank higher; "sales velocity" framing is the industry elaboration. |
| Conversion rate | Industry consensus. Listings that convert clicks into purchases are observed to rank better, consistent with Amazon's goal of surfacing likely sales. |
| Price | Amazon-confirmed as a factor in placement; staying competitively priced is cited in Amazon's own seller guidance as a way to improve ranking. |
| Availability and inventory | Amazon-confirmed. Availability is named as a placement factor; out-of-stock or low-inventory listings lose standing. |
| Customer reviews and ratings | Industry consensus as a ranking lever; Amazon documents reviews as quality signals and offers programs like Vine, but does not state they directly set rank. |
| External traffic | Industry consensus, not Amazon-documented. Practitioners report off-Amazon clicks that convert can strengthen a listing's organic standing. |
| Account and listing health | Industry consensus, partly echoed by Amazon. Seller performance, policy compliance, and listing completeness are reported to influence eligibility and rank. |
| AI-surface fit (COSMO / Rufus) | Amazon-confirmed systems. Content that clearly answers intent and use-case questions is better positioned for search navigation and Rufus recommendations. |
The practical reading: build for the Amazon-confirmed signals first, treat the consensus signals as supporting moves, and never assume an exact weighting Amazon has not published.
How to optimize across all four systems
To rank on Amazon, match real queries for relevance, optimize the whole listing for conversion, keep popular items in stock and competitively priced, use Sponsored Products to seed velocity, write answer-ready content for the AI layer, and drive qualified external traffic.
-
Match real queries in your title, bullets, and backend search terms.
Relevance is the gate Amazon confirms it uses; a listing that does not match the query cannot rank organically or via ads for it.
-
Optimize the listing for conversion, not just keywords.
Amazon states better-selling products rank higher, so images, price, reviews, and copy that turn clicks into sales feed the strongest confirmed signal.
-
Keep popular ASINs in stock and competitively priced.
Availability and price are Amazon-confirmed placement factors; stockouts and uncompetitive pricing directly cost ranking.
-
Use Sponsored Products to seed early sales velocity on new listings.
Ads manufacture the impressions and conversions that become organic signals, which is why ad and SEO management are run together.
-
Write answer-ready, use-case-rich content for the AI layer.
COSMO and Rufus reason about intent and situations, so content that explains who a product is for and when to use it improves AI-surface visibility.
-
Drive qualified external traffic to high-intent listings.
Practitioners report external clicks that convert can lift organic standing; even where unconfirmed, the added sales help the metrics Amazon does reward.
Amazon ranking myths vs. reality
Amazon ranking attracts a lot of confident folklore. Here are the most common myths and what is actually true.
Myth Amazon's current algorithm is officially called A10.
Reality "A10" is seller and agency terminology. Amazon has never confirmed a version change or that name; A9 is the only ranking system tied to a real, named Amazon entity.
Myth More keywords always mean higher rankings.
Reality Relevance gates eligibility, but Amazon states better-selling products rank higher. Beyond a relevant match, conversion and sales history do the heavy lifting, not keyword stuffing.
Myth Rufus runs on Amazon Bedrock, which is the ranking algorithm.
Reality Bedrock is AWS's developer foundation-model service, not Amazon's retail ranking system. Rufus is a separate consumer shopping assistant trained on Amazon's catalog, reviews, Q&As, and web data.
Myth Running ads directly boosts your organic rank.
Reality Ads do not set organic rank directly. They generate sales and reviews, which are organic inputs - the boost is indirect, and Amazon publishes no formula linking spend to rank.
Myth Amazon SEO and Google SEO follow the same playbook.
Reality Amazon ranking is built around transactional intent and purchase likelihood, not informational research. Conversion, price, and availability matter far more than they do on Google.
Spoke guides in this cluster
This pillar is the map. Each system has a dedicated guide:
- The A9 Algorithm Explained - history, relevance vs. performance, and why A9 is the only officially named system.
- The A10 Algorithm Explained - the behavioral factors sellers attribute to today's system, clearly flagged as consensus.
- Rufus and Amazon's AI Search Layer - how COSMO and Rufus change discovery and what to do about it.
- Amazon Sponsored Ads and the Auction - the CPC auction mechanics and how ads feed organic rank.
- Amazon Ranking Factors Checklist - the consolidated optimization list across every system.
Frequently asked questions
A9 traces to A9.com, a real Amazon search subsidiary founded in 2003 whose name is a numeronym for "algorithm." Amazon has never officially confirmed an "A10" version or rename. Treat A9 as legacy shorthand and A10 as seller and agency terminology for the current, more behavior-driven system.
Amazon's own guidance states that search terms, price, availability, selection, and sales history help determine where a product appears, and that better-selling products tend to rank near the top. Everything more granular - exact weightings, external traffic effects - is industry consensus inferred from observed behavior, not published mechanics.
Amazon ranking is built around transactional intent: its goal is to surface products a shopper is most likely to buy. Conversion rate, sales history, price, and availability carry far more weight than on Google, where informational relevance and link authority dominate. On Amazon, a listing that sells well outranks one that merely matches keywords.
Rufus is Amazon's generative AI shopping assistant, launched in beta in 2024. In May 2026 Amazon brought Rufus and Alexa+ together into "Alexa for Shopping," retiring the standalone Rufus brand while keeping its product-knowledge capabilities. It is trained on Amazon's catalog, reviews, community Q&As, and web information to answer shopper questions and recommend products. It shapes discovery, so clear, intent-rich listing content improves your odds of being surfaced.
COSMO is a system documented by Amazon Science that mines common-sense knowledge from shopper behavior and large language models to build knowledge graphs connecting products to intents and use cases. Amazon reports it is deployed in search applications such as search navigation, helping the system understand what a query really means beyond literal keywords.
Sponsored Products, Brands, and Display run cost-per-click auctions that buy visibility. That visibility generates clicks, sales, and reviews, and those are organic ranking inputs. So ads can indirectly strengthen organic position, especially at launch. Amazon publishes no direct formula tying ad spend to organic rank, so the effect is reported by practitioners.
No. Amazon Bedrock is AWS's managed foundation-model service for developers building generative AI apps. It is unrelated to Amazon's retail product-ranking system. The retail-facing AI that affects shoppers is Rufus (the assistant) and COSMO (the search-navigation knowledge graph), not Bedrock itself.
Start with relevance - put real query terms in your title, bullets, and backend search fields - then optimize for conversion with strong images, competitive price, and reviews, since Amazon confirms better-selling products rank higher. Keep popular items in stock, then use Sponsored Products to seed early sales velocity on new listings.
References
- Amazon Science: COSMO common-sense knowledge generation and serving system
- About Amazon: Rufus, Amazon's AI-powered shopping assistant
- Amazon Ads: Sponsored Products overview
- Amazon Seller blog: Amazon SEO guidance for sellers
- Wikipedia: A9.com (Amazon search subsidiary)
- Moz: How to Rank Well in Amazon, the US's Largest Product Search Engine
- Semrush: Amazon SEO - Top Strategies to Optimize Your Product Listings
- Search Engine Journal: The Top Search Engines and AI Search Engines to Know