A9 vs COSMO vs Alexa for Shopping: What Each Amazon System Actually Does
Amazon AI · 8 min read
Amazon does not have one algorithm — it has three overlapping systems that decide whether a shopper sees and chooses your product. A9 ranks the list, COSMO reads intent behind the query, and Alexa for Shopping is the conversational assistant that names a product out loud. Here is exactly what each does and what you should optimize for in 2026.
The short version
A9 — the classic ranking engine. Matches keywords to products and orders them by relevance and conversion.
COSMO — the intent model (Customer Obsession Shaping Model). Understands why someone is searching and clusters results by need.
Alexa for Shopping — the assistant (built on Rufus technology). Takes a natural-language question and recommends specific products by name.
They are layered, not competing. A9 puts you in the candidate pool; COSMO sorts that pool by intent; Alexa for Shopping (or ChatGPT, or Google) quotes the winner. Your content has to serve all three.
What A9 does
A9 is the system most sellers already know. A shopper types "non-slip yoga mat," and A9 returns products containing those terms, weighted by relevance and historical conversion rate. It is keyword-centric: the closer your text matches the query and the better you convert, the higher you rank.
Optimize for A9 by: placing relevant keywords in title, bullets, and backend fields; maintaining conversion rate through pricing, images, and reviews; and keeping your catalog clean.
What COSMO does
COSMO sits beneath A9 and reads intent. For the same query "non-slip yoga mat," COSMO can tell whether the shopper is a beginner (affordable, thick, with a bag) or an advanced practitioner (thin, natural rubber, premium), then clusters results by that need. As of 2026 it is active in 18 categories including Sports & Outdoors, Kitchen, Electronics, and Beauty.
Optimize for COSMO by: describing use cases and buyer context, not just the product term; expanding backend keywords with synonyms and scenarios; and answering typical buyer questions in your bullets so the model can map you to an intent.
What Alexa for Shopping does
Alexa for Shopping (the evolution of the Rufus conversational assistant, consolidated into the Alexa brand in 2026) is the interface shoppers talk to. When someone asks "what is a good yoga mat for a sweaty beginner on a budget," it synthesizes a recommendation — often naming specific products — drawn from your listing content, A+ modules, reviews, and broader web signals. This is where GEO decides whether you get named.
Optimize for Alexa for Shopping by: writing quotable, question-shaped copy; completing A+ content; and keeping brand and attribute facts consistent everywhere the assistant might look.
Side-by-side comparison
Dimension
A9
COSMO
Alexa for Shopping
Role
Ranking engine
Intent model
Conversational assistant
Input
Keyword query
Query + context
Natural-language question
Output
Ordered list
Intent-clustered set
Named recommendation
Core signal
Keyword relevance + conversion
Intent fit + context
Quotable, consistent facts
Where it reads
On-page text
Text + behavior signals
Listing + A+ + reviews + web
Seller lever
Keywords, conversion
Use-case copy
GEO-ready content
How they work together (a worked example)
A shopper asks Alexa for Shopping: "recommend a yoga mat for a beginner on a budget."
A9 retrieves the broad set of yoga mats that match the query terms.
COSMO narrows to the "beginner + budget" intent cluster.
Alexa for Shopping scans the clustered products' copy for clear, quotable claims like "6mm thick, non-slip, includes carry bag, beginner-friendly" and names the one whose content best answers the question.
If your listing only says "Premium Yoga Mat Pro," you survive A9 but get filtered out by COSMO and never quoted by the assistant. If it says exactly who it is for and why, you win at every layer.
What sellers should do differently
Stop writing only for the keyword. A9 still needs keywords, but COSMO and Alexa for Shopping need sentences a human — and a model — can understand.
Answer the question, then sell. Frame each bullet as a response to a buyer question in your category.
Treat A+ as source material. Assistants read it; empty modules are a missed citation.
Keep facts consistent off Amazon too. The same product name and claims across your site and mentions reinforce the entity the assistant trusts.
From understanding to action
You do not have to map all three systems by hand. Our free AI Recommendation Readiness check scores your title and bullets on the signals A9, COSMO, and Alexa for Shopping all reward — clarity, intent, and consistency — and shows the exact lines to improve.
Evidence, not hype: A9, COSMO (18 categories as of 2026), and the Alexa-for-Shopping / Rufus assistant are documented by Amazon and reported by seller sources. Specific interaction details are simplified for clarity; the layered model (rank → intent → recommendation) is the durable mental model.
Frequently asked questions
Q: Does COSMO replace A9? A: No. A9 still ranks the candidate list by keyword relevance and conversion. COSMO works beneath it, clustering results by intent. You need both: A9 to get into the pool, COSMO to be sorted into the right intent cluster.
Q: Is Alexa for Shopping the same as Rufus? A: They are the same assistant lineage. Amazon built the conversational shopping assistant (originally Rufus) and consolidated it under the Alexa for Shopping brand in 2026. You will see both names in seller discussions.
Q: Do I still need keywords if COSMO reads intent? A: Yes. Keywords get you into A9's candidate set in the first place. COSMO and the assistant then decide whether you rise within it. Keywords are the floor, not the ceiling.
Q: Which system should I optimize for first? A: In order: A9 (so you rank at all), then COSMO (use-case copy), then Alexa for Shopping / GEO (quotable, consistent facts). They build on each other; skipping A9 wastes the rest.
Q: How do I make my content quotable for the assistant? A: Write short, specific, question-shaped claims that state who the product is for and why. "Beginner yoga mat, 6mm, non-slip, includes bag" beats vague superlatives. A free readiness check scores this for you.
Q: Does this apply to every category? A: A9 applies everywhere. COSMO is active in 18 categories as of 2026 and expanding. Alexa for Shopping touches all shopping queries over time. Even outside COSMO categories, clear quotable copy helps the assistant recommend you.