Amazon search is no longer only about matching a shopper’s typed keyword to a product.
Amazon says its Rufus shopping assistant answers product questions, comparisons, and recommendation requests using sources that include the product catalog, customer reviews, community Q&A, and relevant Store APIs. Amazon’s research on COSMO also describes a system that connects products with the human contexts in which they are used, including functions, audiences and events.
That creates a practical listing problem.
A product may contain the right keywords and still communicate its use case poorly. A shopper can ask Amazon a question such as “Which of these is better for a small kitchen?” or “Is this suitable for a five-year-old?” The listing needs to make the relevant product facts easy to identify and connect.
Need help optimizing Amazon listings and marketplace content? Visit Alisterr.com.
Table of Contents
- What Rufus and COSMO actually do
- AEO on Amazon: What is proven and what is not
- How to structure the title
- The 5-bullet AEO framework
- Use Item Highlights for use cases and materials
- How to use images without relying on image text
- How to structure A+ Content
- Build semantic relationships into the listing
- AEO-ready listing template
- CRO test plan
- What not to claim about Rufus and COSMO
- Final checklist
What Rufus and COSMO actually do
Rufus is Amazon’s generative-AI shopping assistant. Amazon says it was designed to answer questions about products, comparisons and recommendations. Its underlying system uses retrieval-augmented generation, selecting information from sources that Amazon considers reliable for a particular question. Those sources include the product catalog, customer reviews, community questions and answers, and relevant Stores APIs.
That matters because a conversational query is different from a traditional keyword.
A shopper searching:
“running shoes”
is giving Amazon a product category.
A shopper asking:
“What running shoes are suitable for someone who runs on trails and needs more traction?”
is giving Amazon a product type, use case, audience need and functional requirement.
Amazon’s own research on COSMO describes this type of relationship. Its knowledge graph is designed to connect products with human contexts, including what a product is used for, who uses it and what events or activities it relates to. Amazon gives examples such as connecting a winter coat with warmth-related functionality and slip-resistant shoes with pregnant shoppers.
COSMO is therefore not simply a synonym for “keywords with AI.”
Amazon’s published research says COSMO was built to generate common-sense relationships from customer behavior and was deployed in Amazon search applications including search navigation.
AEO on Amazon: What is proven and what is not
“Answer Engine Optimization” or AEO is a useful industry term for structuring content so an AI system can more easily understand and answer questions from it.
It is not an Amazon-published ranking specification.
Amazon has not published a seller rule saying:
Put this phrase in Bullet 2 and Rufus will recommend your ASIN.
Nor has Amazon published a checklist guaranteeing that a particular A+ module, image caption or bullet sequence will cause Rufus to select a product.
That distinction matters.
There is strong evidence that Amazon is using generative AI and semantic relationships in discovery. There is also direct evidence that Rufus uses Amazon catalog information when answering questions.
There is not public evidence for a fixed “Rufus ranking formula” that sellers can reverse-engineer from five bullet points.
The useful approach is therefore:
Do not optimize for a supposed secret Rufus formula. Optimize the listing so the product’s facts, uses, audience and constraints are explicit and internally consistent.
That is supported by how Amazon describes its systems, while avoiding claims Amazon itself has not made.
How to structure the title
Amazon changed its title requirements in 2026.
Starting July 27, 2026, product titles in categories other than media must be 75 characters or fewer, including spaces. Amazon also introduced Item Highlights, which provide up to 125 additional characters for materials, recommended use cases and other details that help customers compare products. Amazon says Item Highlights are searchable and appear with titles in search results and on product detail pages.
This creates a useful division of information.
The title should establish the product clearly.
Item Highlights can carry an additional concise use case, material or differentiating detail.
Title template
Use this as a writing framework, not an Amazon-published AEO formula:
[Brand] + [Product Type] + [Core Differentiator] + [Key Variant/Specification]
Example:
NorthPeak Insulated Lunch Bag, Leakproof, 12-Can Capacity
The goal is not to cram every possible search term into 75 characters.
Amazon says titles should help customers understand the product, and its new title workflow is designed to keep key product information in the title while moving additional details into Item Highlights.
That makes a concise, fact-dense title more defensible than a title built around repetitive keyword strings.
What to avoid
Do not sacrifice product clarity to fit more synonyms.
Do not repeat the same noun multiple times.
Do not turn the title into a sentence containing every possible use case.
And do not treat the 75-character limit as an AEO trick. It is an Amazon title requirement.
The 5-bullet AEO framework
Amazon says bullet points should communicate features and benefits clearly and concisely. Amazon’s seller guidance recommends starting with a feature and then stating its benefit, providing product-specific data, and putting bullets in order of importance.
Amazon also stated that its listing-quality process uses generative AI to help optimize bullet-point quality and can generate compliant bullet points for seller review.
There is no Amazon-published “Rufus five-bullet sequence.”
But you can create a useful sequence from Amazon’s documented emphasis on product facts, benefits, use cases and customer decision-making.
Bullet 1 — What it is
Define the product and its most important differentiator.
Template:
[Core feature] — [what the product is] for [primary user/use]
Example:
12-Can Insulated Capacity — Holds up to 12 standard cans and keeps meals organized for work, school or travel
Bullet 2 — Primary function
State what the product actually does.
Template:
[Functional feature] — [specific customer benefit]
Example:
Leak-Resistant Lining — Helps contain spills from drinks and prepared meals during daily transport
Bullet 3 — Main use case
Connect the product to a real situation.
Template:
Designed for [specific use case] — [relevant product capability]
Example:
Workday Meal Storage — Separate lunch, snacks and drinks in one compact bag for office or commuting use
Bullet 4 — Specification or constraint
Give the information a comparison shopper would need.
Template:
[Dimension/material/compatibility/specification] — [practical implication]
Example:
10 × 8 × 6-Inch Size — Fits easily in most commuter bags while providing dedicated food storage
Bullet 5 — Important objection or requirement
Use the final bullet for information that can prevent uncertainty.
Template:
[Care/compatibility/limitation] — [what the customer needs to know]
Example:
Easy-Clean Interior — Wipe the insulated lining clean after everyday use
This sequence is an operational framework, not an Amazon ranking rule.
Its purpose is to make five bullets collectively answer five basic questions:
What is it? What does it do? When would I use it? What are the important specifications? What else do I need to know?
That structure is consistent with Amazon’s guidance to make bullets clear, product-specific and decision-oriented.
Use Item Highlights for use cases and materials
Amazon’s new Item Highlights field is particularly useful for concise contextual information.
Amazon says the field provides an additional 125 characters for materials, recommended use cases and details that help customers compare products, and that it is searchable and visible in search results and on product detail pages.
That makes Item Highlights different from hidden backend search terms.
A useful framework is:
Material + use case + meaningful differentiator
Example:
Stainless steel body; designed for daily commuting; fits standard cup holders
The example is illustrative. The actual content should come from verified product specifications.
Do not use Item Highlights to make claims that the product cannot substantiate.
How to use images without relying on image text
Images still matter for customers, but this is where sellers should avoid one of the biggest unsupported AEO claims.
Amazon’s public technical description of Rufus identifies the product catalog, customer reviews, community Q&A and Stores APIs as information sources. It does not publish a statement saying:
“Rufus reads text embedded in your listing infographics and gives that text ranking weight.”
That means sellers should not make an infographic the only place where an important product fact appears.
Put critical information into the listing’s textual product data as well.
For example, do not make your image say only:
“BUILT FOR SMALL SPACES”
and leave the actual supporting fact inside the graphic.
Instead, the listing text should state what makes it suitable:
“14-inch width fits compact kitchen counters.”
The image can then reinforce the same fact visually.
Image microcopy framework
Use short, customer-readable text such as:
Feature → measurable fact → use
Examples:
Foldable → 4-inch storage height → easy cabinet storage
USB-C → 65W charging → suitable for compatible laptops
Water-resistant → IPX4 rating → designed for light splashes
The exact claims must come from the product.
The principle is simple:
The image supports the claim. The text field states the claim.
That gives customers a visual explanation without depending on an unverified assumption about Amazon’s image-processing system.
How to structure A+ Content
Amazon describes A+ Content as supplemental marketing content that can explain product features in different ways. Basic A+ complements the bullets and main images, while Premium A+ adds richer modules, videos, comparison charts, carousels and a Q&A module. Amazon explicitly says A+ gives brands an opportunity to answer customers’ common questions with relevant product and brand information.
That makes A+ useful for depth, even though sellers should not describe individual modules as guaranteed Rufus ranking signals.
A practical A+ order is:
Module 1: Product problem
State the problem the product solves.
Module 2: Core function
Explain how the product addresses it.
Module 3: Use cases
Show the major contexts in which the product is intended to be used.
Module 4: Feature proof
Break important features into specific facts, measurements or compatibility information.
Module 5: Comparison
Explain differences among products in the same brand where relevant.
Module 6: FAQ/Q&A
Address legitimate questions customers commonly have.
This order is a content strategy, not an Amazon-published “Rufus module ranking.”
It works because the customer can progressively move from:
problem → product → function → use → evidence → comparison → questions
without forcing every fact into one image or one bullet.
Amazon’s own A+ guidance says the content should complement the bullets and images and proactively answer customer questions.
Build semantic relationships into the listing
This is where COSMO provides a useful technical clue.
Amazon’s COSMO research describes relationships such as:
used_for_function
used_for_event
used_for_audience
and other common-sense connections between products and customer contexts.
A seller cannot submit a COSMO relationship manually.
But the research suggests a practical content discipline:
Don’t describe only what the product is.
Describe:
who it is for
what it is used for
when it is used
what problem it solves
what it is compatible with
what differentiates it
For example, instead of:
“Portable projector with 1080p resolution.”
a more informative listing might state:
“1080p portable projector designed for movie nights, presentations and compact rooms; supports [verified compatibility].”
That adds relationships among:
product → resolution → use case → environment → audience
Again, this is a content implication from Amazon’s published research, not a promise that writing this sentence will make an ASIN rank higher.
AEO-ready listing template
Use this framework when rewriting a listing.
Title
[Brand] [Product Type] [Primary Differentiator] [Key Specification]
Keep within Amazon’s applicable title limit.
Item Highlights
[Material or specification]; [primary use case]; [meaningful comparison detail]
Bullet 1 — Product identity
[Feature/specification] — [what the product is and who it is for]
Bullet 2 — Function
[Feature] — [what it does and the measurable/practical benefit]
Bullet 3 — Use case
Designed for [specific situation] — [relevant capability]
Bullet 4 — Comparison fact
[Size/material/compatibility/specification] — [why this matters when comparing options]
Bullet 5 — Requirement or objection
[Care/compatibility/limitation] — [important information before purchase]
Image 1
Show the product clearly.
Image 2
Show the primary use case.
Image 3
Show a measurable feature or specification.
Image 4
Show compatibility, dimensions or components.
A+
Problem → solution → features → use cases → comparison → FAQ
This is not a secret formula for Rufus.
It is a way to prevent important product information from being scattered, duplicated or left only inside graphics.
CRO test plan
The biggest mistake is changing the listing and then claiming that an increase in sales proves the listing became “Rufus optimized.”
Amazon has not published a seller metric showing that an ASIN was selected by Rufus because of a specific bullet, title or A+ module.
Test the actual business outcome instead.
Test 1: Listing clarity
Compare the current listing against the rewritten version.
Track:
- organic sessions
- conversion rate
- click-through rate where available
- unit session percentage
- returns
- customer questions
Test 2: Question coverage
Create a list of 10–20 real customer questions.
Examples:
Is this suitable for [audience]?
Will it fit [specific environment]?
Is it compatible with [device]?
What is it made from?
How much does it hold?
Then check whether each question has a direct, verifiable answer somewhere in the listing.
Test 3: Rufus observation
Where Rufus/Alexa for Shopping is available, ask the same product questions before and after the content change.
Keep the prompt wording consistent.
Record:
Question → products returned/recommended → facts mentioned → inaccuracies → date
Do not treat one answer as proof of a ranking change. Amazon says Rufus uses multiple evidence sources, and the relevance of those sources varies by question.
Test 4: Conversion test
If you have sufficient traffic, compare the old and new listing versions using a controlled methodology.
The primary business question is:
Did clearer product information help customers choose the product?
That is a stronger test than trying to infer a secret AI score.
What not to claim about Rufus and COSMO
“Rufus only reads bullets”
Amazon does not say this.
Amazon says Rufus uses the product catalog, reviews, community Q&A and relevant Stores APIs.
“A+ Content directly controls Rufus ranking”
There is no public Amazon statement supporting that claim.
Amazon describes A+ as supplemental product information and a way to answer customer questions, but it does not publish a Rufus ranking weight for A+.
“Text inside images is an Amazon AI ranking factor”
Amazon’s published Rufus architecture does not establish this.
Use image text to help customers, but repeat important specifications in textual listing content.
“COSMO replaces keywords”
Amazon’s COSMO research does not say that.
It describes a common-sense knowledge system used in Amazon search applications. Keyword matching and other retrieval systems can still exist alongside semantic systems.
“There is a fixed five-bullet Rufus sequence”
There is not an Amazon-published sequence.
The five-bullet framework above is a practical content structure based on Amazon’s own guidance around product-specific information, features, benefits and customer decision-making.
“AEO guarantees AI recommendations”
It does not.
Amazon’s AI systems use multiple information sources and, in Rufus’s case, retrieval relevance varies by question.
Why keyword stuffing is the wrong response
Amazon’s own bullet guidance says well-written bullets should communicate clearly and naturally contain relevant keywords, rather than treating keywords as the primary objective. Amazon recommends clear, concise bullets focused on important product information.
That fits the broader direction of Amazon’s AI systems.
COSMO was designed to capture relationships such as audience, function and event, while Rufus was designed to answer natural-language shopping questions from multiple evidence sources.
A listing that says:
“water bottle bottle sports bottle gym bottle reusable bottle…”
contains repeated words.
A listing that says:
“24-ounce insulated stainless-steel bottle designed for gym bags, commuting and daily hydration; fits standard cup holders”
contains more usable product relationships.
It identifies:
product → size → material → use case → audience/context → compatibility
That is more informative to a customer and more consistent with the kind of contextual understanding Amazon describes in its AI research.
Final checklist
Before publishing an Amazon listing intended to work well for both conventional search and AI-assisted discovery, check:
Title: Is the product immediately identifiable?
Item Highlights: Do they add useful material, use-case or comparison information?
Bullets: Does each bullet add a distinct, verifiable fact?
Sequence: Are the most important customer decisions covered first?
Semantics: Does the listing state who the product is for, what it does and when it is used?
Specifications: Are important dimensions, materials, compatibility details and limitations explicit?
Images: Do visuals reinforce the claims instead of being the only place those claims appear?
A+: Does the content answer real customer questions rather than repeat the title five times?
Consistency: Do the title, bullets, images, A+ and attributes agree?
Accuracy: Can every factual claim be supported by the actual product?
Testing: Are you measuring conversion and customer outcomes rather than assuming a change improved Rufus visibility?
Final takeaway
Amazon’s AI discovery systems create a good reason to make listing information more structured, but sellers should separate evidence from AEO marketing claims.
Amazon says Rufus uses the product catalog, reviews, community Q&A and relevant Store data to answer customer questions. Amazon’s COSMO research shows how Amazon can model relationships between products and customer contexts such as audiences, functions and events.
That does not mean sellers can manipulate Rufus with a prescribed keyword formula.
The practical opportunity is simpler.
Write listings that clearly identify the product, explain what it does, state who it is for, describe relevant use cases, provide concrete specifications and answer the questions that could determine a purchase.
Use titles for product identity. Use Item Highlights for concise additional context. Use bullets for structured product facts. Use images to demonstrate those facts. Use A+ to provide depth and answer questions.
The result is not just a listing designed around keywords.
It is a listing with enough clear product information for both customers and Amazon’s increasingly AI-driven shopping systems to understand what the product is, who it fits, and when it makes sense.
Need help with Amazon SEO, A+ Content, listing optimization, or marketplace content? Visit Alisterr.com.
Leave a Reply