Alexa for Shopping and COSMO: How to Get Your Products Recommended on Amazon
Tutorials & Tips
18 min1 min read
Alexa for Shopping leverages COSMO to decode buyer intent and curate top product recommendations from millions of SKUs. This guide outlines the 60 optimization ranking factors to boost your Amazon PDP visibility and conversion.
Alexa for Shopping and COSMO: 60 Amazon Ranking Factors
On May 13, 2026, Amazon retired the Rufus brand, merging its AI engine into a unified assistant: Alexa for Shopping. Unlike Rufus, which lived in a separate chat window that few shoppers ever opened, Alexa for Shopping is now integrated directly into Amazon's main search bar. Every shopper encounters it, whether they ask for it or not.
Under the hood, COSMO is still driving the experience. COSMO is the knowledge graph that understands the intent behind shopper queries. When someone searches for "wedding shoes," COSMO doesn't just look for the words "shoes" and "wedding" in the listings. It identifies the intent—elegance, comfort for a long day, formal dress code—and surfaces products that meet this intent, even if they don't contain those exact keywords.
The central question for any brand selling on Amazon in 2026 is simple: how do you get surfaced by this AI? AI responses can only display 3 to 5 products, compared to 20 to 40 in traditional search results. The competition for these spots has intensified.
This article is a comprehensive guide. It covers the 60 identified criteria that Alexa for Shopping and COSMO use, grouped by category, with exact thresholds verified as of July 2026.
Table of Contents
Section 01
The 2026 Context: Alexa for Shopping, COSMO, and Agentic AI
Amazon has fundamentally changed how shoppers discover products. Before diving into specific criteria, you must understand the three concepts governing Amazon visibility today.
Alexa for Shopping (formerly Rufus)
Alexa for Shopping is Amazon's conversational AI assistant that answers buyer queries directly in the search bar. When a shopper inputs "what headphones for working from home with screaming kids," they no longer just get a list of results—they receive a conversational response with 3 to 5 recommended products.
This shift is strategic. On a classic SERP, a shopper might scroll down to the 40th product. In an AI response, they only see 3 to 5. Securing a spot in these 3-5 recommendations is now priority number one.
COSMO—The Semantic Engine
COSMO (COmmon Sense knOwledge) is the knowledge graph powering Alexa for Shopping. It doesn't rely on simple keyword matching—it focuses on intent understanding. It analyzes 1.87 million "search-to-purchase" pairs and 3.14 million "co-purchase" pairs to understand what a shopper actually wants when they type a query.
In practice, COSMO prioritizes listings that describe use-case contexts, user personas, and resolved pain points. It penalizes listings stuffed with keywords that lack narrative flow.
A9, A10: The Traditional Search Engine
A9 (sometimes called A10 in the seller community, though Amazon has never officially named "A10") is Amazon's traditional search engine. It indexes keywords, measures frequency, position, and click-through rate (CTR) to rank products. It remains active in 2026: most shoppers typing simple queries ("smart scale") still use A9 to view the traditional search results page. COSMO acts as an AI layer on top of A9.
The Strategic Takeaway
A product listing must now appeal to both systems: A9 to secure traditional SERP visibility, and COSMO to be recommended by Alexa for Shopping. The two algorithms operate on different logic—which is where the complexity lies.
The Amazon Quality Benchmark
Since 2026, Amazon has leveraged its Listing Quality Dashboard as an active ranking signal. A product whose quality score drops below 65% faces organic ranking suppression. This dashboard is visible in Seller Central, but its calculation remains opaque, quietly penalizing millions of listings.
Section 02
Product Title
The title is the first element analyzed by every system: A9 for indexing, COSMO for product understanding, and the shopper for click decisions. It is governed by six criteria.
Title Length
The optimal title length in 2026 is 180 to 200 characters. Titles that are too short miss indexing opportunities, while overly long titles degrade on mobile.
Important Note: Amazon announced that starting July 27, 2026, the limit will drop to 75 characters across all categories except Media, introducing a new 125-character "Item Highlights" field to compensate. This rule is confirmed for the US market but is not yet active in Europe. We are monitoring its global rollout.
Attribute Formatting
COSMO heavily favors front-loading: the first 80 characters must feature the brand, core keyword, and primary differentiator. This is critical for mobile optimization and is the primary data Alexa analyzes to identify the product.
Brand-First Structure
Titles must start with the brand name. This is an official Amazon policy, and COSMO uses the brand as a semantic anchor for all conversational recommendations ("I'm looking for a [product] from Brand X").
Permitted Characters
Amazon strictly prohibits the following characters in titles unless they are part of a brand name: ! $ ? _ { } ^ ¬ ¦. Others, like ~ # < > *, are only accepted in specific contexts such as identifying a style (Style #131) or a measurement (>3kg). The characters + = & ( ) / , . : - are permitted.
Zero Promotional Terms
Words like "promo," "best price," "best seller," "#1," "free," and equivalents are banned in titles. Amazon can suppress listings from search results for non-compliance. These terms hold zero semantic value for COSMO—they waste valuable character count without driving visibility.
Semantic Richness for AI
This is the most critical factor for COSMO. An AI-optimized title contains 6 distinct elements:
Brand Name
Product Name
Core Function (what the product does)
Target Audience or Use Case (who it is for, under what conditions)
Key Differentiator (material, technology, size)
Format or Specification (color, quantity, pack size)
A title covering all 6 elements is highly indexable by COSMO across a wide range of conversational search queries. Titles covering only 2-3 are restricted to highly specific queries.
Common Pitfall
Keyword stuffing (e.g., piling synonyms like "Insulated Water Bottle, Thermos, Hot/Cold Flask" into the title) actively harms AI visibility. COSMO detects and penalizes this. Use one primary keyword—synonyms belong in backend search terms (see below).
Section 03
Bullet Points
Bullet points are the core of your product listing. This is where COSMO extracts the majority of its intent signals. They are governed by ten criteria.
Number of Bullet Points
Amazon recommends at least 5 bullet points. Listings with fewer than 3 lose structural hierarchy, leaving the AI with less surface area to extract signals. Exceeding 5 is not penalized, but extra content won't boost visibility once the indexing limit is hit.
Bullet Length
Each bullet should be between 180 and 250 characters. This is the sweet spot for mobile readability and COSMO's semantic indexing.
Byte Indexing Limits
Amazon only indexes the first 1,000 cumulative bytes of your bullet points. Any text beyond this remains visible to shoppers but is ignored by both A9 and COSMO.
Critical Detail: Bytes do not equal characters. Special or accented characters (é, à, ç) cost 2 bytes, and emojis cost 4 bytes or more. In non-English listings, you will hit the byte limit much faster than character counts suggest. "Heat-resistant" might be 14 characters, but can consume more bytes depending on formatting.
Cumulative Volume | Indexing Impact |
|---|---|
0-1,000 bytes | Fully indexed |
1,001-1,200 bytes | Slight overage, trailing content ignored |
1,201-1,500 bytes | Significant portion not indexed |
Over 1,500 bytes | Majority of content invisible to AI |
Use-Case Signals
This is COSMO's primary focus. Every bullet must answer one of three questions: who is it for, in what context, or in what situation. These signals can be brief (e.g., "for creators," "for the office," "on-the-go") or highly descriptive (e.g., "engineered for hypertensive adults monitoring blood pressure at home"). Both formats are crawled.
Compatibility and Ecosystem Signals
COSMO analyzes "co-purchase" patterns (mapping 3.14 million purchase pairs). Products that explicitly state compatibility (e.g., "syncs with Health Mate app," "compatible with iOS and Android," "best paired with an athletic armband") are integrated into cross-recommendation algorithms like "customers also bought" or "works well with." Listings lacking these compatibility signals are omitted from those suggestions.
Frequency and Usage Timing
COSMO prioritizes explicit temporal anchors: "every morning before breakfast," "pre- and post-workout," "during your monthly check-up." These anchors help the AI map the product to shopper routines. Without them, products won't surface for routine-based queries ("best product for morning routines," "what to take post-workout").
Environmental Context
Different from use-case signals (which focus on the user), environmental context describes where and under what conditions the product is used: living room, office, outdoors, travel, beach, mountains. COSMO maps these environments to match queries like "best product for [environment]."
Value-Linked Differentiation
A valid product differentiator must be quantified, certified, or comparative. Subjective adjectives ("high quality," "ultra-durable") are ignored. Isolated metrics ("500g") have low weight. A metric linked directly to a benefit ("500g lightweight design for all-day fatigue-free use") or a comparison ("30% lighter than leading competitors") represents the strongest format, known as BENEFIT LINKED.
Technical Specifications
Bullet points must feature at least 3 precise technical specs (dimensions, weight, materials, compatibility, capacity). Amazon leverages this data for long-tail search indexing and answering buyer questions via Alexa for Shopping.
Problem-Solution Mapping
COSMO favors listings that explicitly state the pain point they resolve: "prevents data loss," "relieves lower back pain," "protects against power outages," "stops leaks." Direct problem-solution phrasing allows the AI to answer search queries framed as problems ("my back hurts," "how to prevent data loss").
Section 04
Product Description
The description offers more space than bullet points, allowing you to build a narrative and address more nuanced buyer questions. It is governed by four criteria.
Description Length
The optimal product description length is between 1,000 and 2,000 characters. Under 1,000 characters, the listing lacks the semantic depth needed for COSMO to map out use cases. Beyond 2,000, returns diminish and the AI de-prioritizes the excess length.
Narrative Structure
COSMO values descriptions that follow a clear narrative flow: use-case scenario → pain point → solution → concrete benefit. Descriptions that read like a dry list of specs are penalised, while those that show how the product fits into a customer's daily routine score highly.
Answering the 7 Core Buyer Questions
An optimized description must address the 7 questions 2026 buyers ask:
What exactly is this product?
Who is it designed for?
In what scenarios should it be used?
What specific problem does it solve?
How does it stand out from alternatives?
How do you actually use it?
What is included in the box?
xIntent—Cross-Category Intent Mapping
COSMO tracks "xIntent"—intents that span multiple product categories. A buyer searching for "how to improve my sleep" might need a mattress, a weighted blanket, an essential oil diffuser, or a smart light. Descriptions that explicitly address these cross-category intents allow your product to surface for broad lifestyle queries.
Section 05
Backend Keywords (Search Terms)
This is arguably the most under-optimized indexing asset on Amazon. It features one single, critical compliance threshold.
Search Terms Volume and Format
Backend keywords are hidden from buyers but configured in Seller Central or Vendor Central. They allow A9 to index your product for search terms that don't appear in your title, bullets, or description. In short: index your product for real terms buyers use, without cluttering your public-facing copy.
Field Length | Optimization Rating |
|---|---|
200-249 characters | Optimal |
150-199 characters | Under-optimized |
50-149 characters | Weak |
Under 50 characters | Virtually empty |
Over 249 characters | COMPLETE field de-indexing |
The Absolute Rule of Backend Indexing
Amazon strictly limits this field to 249 bytes. Going even one byte over will cause the entire field to be ignored—not just the extra words, but the whole set. Separate search terms with single spaces (no commas), do not repeat words already used in your title or bullets, and never include competitor brand names.
Section 06
Product Imagery
COSMO is multimodal: it analyzes images with the same depth as text. We score visual optimization across six core criteria.
Pure White Background Main Image
The main image must feature the product on a pure white background, filling at least 85% of the frame, with no overlay text or decorative logos. This is an official Amazon policy. Non-compliant main images risk automatic listing suppression.
Multi-Angle Secondary Packshots
COSMO analyzes multiple views to build a 3D semantic understanding of the product. Best practice requires 3 secondary angles beyond the main image: a 3/4 view, a profile view, and a back view. Without these, the AI's product recognition is limited, restricting its ability to describe and compare your item.
Active Lifestyle Imagery
A lifestyle image must show a real person actively using the product within a clear, recognizable setting. It shouldn't be a white-background packshot or a diagram. This is critical because COSMO reads images to match products with real-world use cases. A model simply posing next to the product does not qualify—the active interaction must be visible.
Diverse Environmental Contexts
If your product can be used in multiple spaces (e.g., living room, office, outdoors), showing these diverse contexts in your secondary images broadens COSMO's recommendation potential. Products shown in only one environment are restricted to highly specific contextual searches.
Infographics and Visualized Value
Infographics that translate technical specs into clear visuals (annotated dimensions, comparative benefits, compatibility charts) are highly valued by COSMO. They allow the AI to visually cross-reference and verify your listing's text claims.
Product Video
Including a product video is a strong quality signal. It shows the product in action and increases on-page dwell time—a performance metric that A9 rewards through better conversion rates. While Amazon does not make video mandatory, listings without it are underrepresented in Alexa for Shopping recommendations.
Section 07
Social Proof
Ratings, reviews, and distribution are the core trust signals COSMO uses to make recommendation decisions. Verified thresholds are outlined below.
Average Rating
This is the most visible metric, yet its direct algorithmic thresholds are rarely documented. Amazon's actual 2026 thresholds are:
For A9 (Traditional Search Rankings)
Average Rating | Ranking Impact |
|---|---|
4.5 stars and above | Optimal zone: triggers a 15% to 25% ranking boost |
4.3 - 4.49 stars | Acceptable baseline for page 1 placement |
4.0 - 4.29 stars | Triggers a 20% to 40% rank suppression |
Below 4.0 stars | Severe rank degradation |
For Alexa for Shopping (AI Recommendations)
Average Rating | AI Recommendation Impact |
|---|---|
4.5 stars and above | Baseline standard for recommended products |
4.3 - 4.49 stars | Eligible for recommendation but not prioritized |
4.0 - 4.29 stars | Highly restricted recommendations |
Below 4.0 stars | Near-total exclusion from Alexa for Shopping |
Review Count
Review volume signals product maturity. Under 10 reviews, a product is considered unverified. Between 10 and 49, trust remains volatile. At 50+ reviews, COSMO can identify recurring strengths and weaknesses to synthesize a reliable recommendation. At 500+ reviews, a brand establishes a trust plateau that acts as a strong competitive moat.
Positive Review Ratio (4-5 Stars)
Average ratings don't show the full picture. A 4.3-star product might have 80% 5-star ratings and 20% 1-star ratings, or a flat distribution yielding the same average. COSMO evaluates this distribution pattern, not just the average. A strong 5-star ratio (over 75%) enables the AI to highlight high customer satisfaction in its conversational responses.
Negative Review Ratio
The ratio of 1-2 star reviews must remain below 5%. Beyond this, COSMO will actively call out recurring defects in conversational replies when buyers inquire about the product. A high negative review ratio is not just a passive conversion drag—it becomes an active negative signal the AI shares with users.
The Operational Reality
A listing with a 4.3 rating and a 3% return rate will be rewarded by Amazon's systems, while a 4.8-rated listing with a 15% return rate will face ranking penalties. Star ratings alone are not enough—they must align with post-purchase behavior.
Section 08
Amazon Badges
Badges are external trust signals awarded automatically by Amazon. In 2026, Alexa for Shopping weights the Amazon's Choice badge roughly twice as heavily as before—making it a core strategic objective.
The Amazon's Choice Badge
Amazon's Choice is awarded programmatically; there is no application process. The badge is awarded per search query: a single ASIN can hold the Choice badge for "sports bluetooth headphones" but not for "bluetooth headphones." Five conditions must be met simultaneously:
Average rating of 4.0+ stars (4.3+ is the optimal target)
Active Prime Status (via FBA, SFP, or highly optimized FBM)
Consistent, reliable stock levels
Return rate below the category average
Conversion rate of 8% to 12% on the target query
The badge is only as strong as its weakest link. A product with a 4.6 rating, solid stock, Prime status, and a low 4% return rate will miss the badge if its conversion rate on the target query is only 6%.
Alexa for Shopping's Impact on Amazon's Choice
Since May 2026, Alexa for Shopping has treated the Amazon's Choice badge as a primary trust signal. Observed weighting is nearly double what it was in previous versions. When selecting 3 to 5 products to recommend, Alexa prioritizes Amazon's Choice listings first, only presenting non-badged alternatives if Choice items do not align with the shopper's specific request.
The Best Seller Badge
Awarded to the #1 product in its specific category or sub-category. It requires high, sustained sales velocity and continuous stock (72 hours of out-of-stock status can trigger badge loss, requiring 4 to 6 weeks to recover).
The Prime Badge
Automatically granted to FBA inventory. For SFP merchants, maintaining this badge in 2026 requires meeting a 93.5% on-time delivery rate and keeping order cancellations under 2%.
Prime Eligibility AP-01
The Prime badge guarantees placement in Prime-filtered searches, which represent the majority of transactions. Eligibility must remain stable at over 90% over a rolling 30 days. Fluctuation degrades ranking and disqualifies products from Amazon's Choice.
Overall Pick, New Release, and Small Business Badges
Overall Pick is similar to Amazon's Choice but targets broader, head-term queries. New Release is reserved for newly launched products with high initial sales velocity. Small Business is a self-declared, free badge for companies under 100 employees that drives a 13.5% conversion lift with zero extra effort.
Section 09
The Buy Box and Commercial Offer
Securing the Buy Box is non-negotiable for success on Amazon. It is governed by three factors.
Buy Box Ownership
The Buy Box (officially renamed the Featured Offer as of 2026) is the default "Add to Cart" option. Listings without the Buy Box are effectively invisible; shoppers rarely click "See All Buying Options." Buy Box share below 80% severely hurts listing performance.
Major 2026 Buy Box Update
In early 2026, Amazon removed seller performance metrics (such as Order Defect Rate) as a primary gatekeeper for Featured Offer eligibility. Buy Box competition has intensified, with the algorithm arbitrating between sellers on price and delivery speed with much faster latency.
Buy Box Stability
Winning the Buy Box is only the first step; maintaining it is what matters. A Buy Box that frequently rotates to third-party sellers is flagged as an unstable offer, which COSMO factors into its recommendation logic. Stability is measured on a rolling 30-day window.
Condition: New
If the Buy Box is captured by a used offer ("Used - Like New," etc.), organic rankings for both A9 and COSMO are degraded. This suggests stockouts from the brand owner, allowing third-party resellers to take over.
Section 10
Availability and Pricing
Out-of-stock or poorly priced items will not be recommended by AI, regardless of listing quality. These are governed by six factors.
Instant Stock Status
A product must be in stock at the exact moment of search scoring. Even brief stockouts cause immediate ranking drops and can trigger the loss of the Amazon's Choice badge.
30-Day In-Stock Rate
Amazon tracks stock continuity over a 30-day window. Maintaining a 95%+ in-stock rate is optimal. Dropping below 90% flags supply chain instability, causing the algorithm to de-prioritize the ASIN.
Price Stability
A stable pricing history is a strong trust signal. Frequent, erratic price hikes can trigger Buy Box suppression if the offer exceeds Amazon's "competitive price" threshold. Focus on price consistency over a 30-day window, avoiding fluctuations greater than 10%.
Price Consistency Across Variations
Massive price discrepancies across variations of the same product line create buyer friction and confuse A9. Coherent, logical variation pricing boosts buyer trust and conversion rates.
Prime Eligibility Longevity
As noted in the badges section, keeping your Prime eligibility above 90% over a 30-day window is mandatory to preserve visibility in Prime-filtered searches.
Section 11
Buyer Guarantees
Three key criteria ensure that a product can be purchased, returned, and used with minimal friction.
Complete Technical Attributes
Amazon rewards listings that populate every structured attribute field in the backend console (dimensions, materials, compatibility, weight, certificates). These fields feed directly into COSMO to answer specific buyer queries ("is this compatible with X," "what are the exact dimensions"). Listings with fewer than 10 populated attributes lack the data depth required by conversational AI.
Category Placement Accuracy
Products must be classified in their correct browse nodes. Misclassified products face search penalty restrictions and are excluded from Alexa for Shopping recommendations for queries targeting their true category.
Non-Add-On Status
Add-on items cannot be purchased on their own—shoppers must hit a minimum order threshold to buy them. This friction destroys conversion rates and excludes the product from most Alexa for Shopping recommendations. Amazon automatically applies Add-on status based on price; a slight price adjustment is often enough to lift this classification.
Section 12
What Else COSMO Analyzes
Beyond traditional A9 ranking factors, COSMO leverages unique user-interaction signals. These metrics are difficult to track externally but are critical for Alexa for Shopping visibility.
External Traffic to Amazon
COSMO rewards listings that drive high-intent traffic from Google, social media, newsletters, and influencers—provided that traffic converts. Driving traffic that bounces harms your listing. High-converting external traffic is heavily weighted; it serves as a strong market validation signal independent of Amazon's internal search.
Review Recency
A product with 300 reviews from 2023 is flagged as historically relevant but currently stagnant. A product with 15 reviews from the last 60 days is flagged as highly relevant now. COSMO values recent review velocity far more than historical volume.
Brand Registry Status
Active Brand Registry enrollment is a fundamental trust signal. COSMO inherently favors registered brands in conversational recommendations. While not strictly mandatory, it acts as a significant differentiator.
Return Rate
Under margin pressures, Amazon has increased the algorithmic weight of return rates. A return rate 2x to 3x higher than the category average can trigger the immediate loss of the Amazon's Choice badge and de-prioritization in AI recommendations.
The Product Subtitle Field
Amazon introduced a dedicated "subtitle" field displayed directly below the main title on mobile. This provides extra semantic real estate without cluttering the main title. COSMO crawls the subtitle as a distinct, valuable indexing field that remains under-optimized by most sellers.
AI Review Summaries (customers_say_summary)
In 2026, Amazon began programmatically generating AI review summaries on detail pages. This summary is synthesized by their proprietary LLM based on recent reviews, capturing both product highlights and defects. This text serves as a direct input for COSMO when evaluating your listing.
Frictionless Path to Purchase
COSMO factors purchase friction into its recommendation algorithms. Products with zero transactional friction (in stock, Prime-eligible, stable Buy Box, non-Add-On) are recommended across all contexts. Listings with transactional friction are de-prioritized, even if their content score is perfect.
Section 13
Action Plan: Where to Start
Optimizing for all 60 criteria at once is impossible. Follow this prioritized roadmap to drive maximum impact with optimized effort.
Priority 1—Resolve Hard Blockers
Immediately fix issues that exclude your product from recommendations: resolve Add-On status, pull ratings above 4.0 stars, win back the Buy Box, stabilize Prime eligibility, and correct browse nodes. These are binary "on/off" switches—no other optimization efforts will yield results until these are resolved.
Priority 2—Optimize Title and Backend Search Terms
These are your fastest levers for indexing growth. Ensuring your title covers the 6 core elements (brand, product, function, audience, differentiator, format) and maximizing search terms to 200-249 bytes can double your indexing footprint within hours.
Priority 3—Inject Bullet Points with Use-Case Signals
Revise bullet points to feed COSMO: specify the user persona, usage context, timing, solved pain points, and product compatibility. This is the single most effective way to gain visibility within Alexa for Shopping.
Priority 4—Deploy Lifestyle Imagery
If your listing only features white-background packshots, you are invisible to COSMO for contextual searches. Adding 2 to 3 lifestyle images showing a person using the product in a real-world setting offers a high return on investment.
Priority 5—Drive Recent Review Velocity
COSMO favors fresh customer feedback. Regularly trigger the "Request a Review" feature in Seller Central and enroll new products in the Amazon Vine program to maintain a consistent stream of recent reviews.
The Core Takeaway
None of these factors operate in a vacuum. A product with a perfect title but a 3.8-star rating will not be recommended. A product with a 4.7-star rating but a title that targets no search intent will not be found. Alexa for Shopping visibility is driven by the holistic combination of these criteria, not isolated optimizations.
Book a meeting with our experts.
A 303 session to discuss your supply chain, finance, or AI visibility challenges.
Related Products
Similar Products
