Your best-performing Amazon listing just lost 15% of its impressions. No ad changes. No pricing shift. No new competitor. The listing that replaced yours in the top five has half the reviews and a fraction of the ad spend — but its product description reads like a direct answer to a question a real shopper would ask. That's not a fluke. That's COSMO at work. And if your brand manager or agency is still optimizing titles by shuffling keyword order, you're optimizing for a system that's already been replaced.
COSMO (Common Sense Knowledge for E-commerce) is Amazon's LLM-based product ranking system that uses commonsense knowledge graphs and semantic understanding to match products with shopper intent — replacing the keyword-matching logic of the legacy A9 algorithm. It doesn't reward you for stuffing "organic dog treats grain-free all-natural" into your title six ways. It rewards you for describing a product that actually answers what a shopper looking for grain-free organic dog treats needs to know.
This guide is for brand managers, eCommerce leads, and CPG operators who manage product listings on Amazon and need to understand what changed, why it changed, and what to do about it. We'll cover what COSMO actually does, the four signals it reads, the Rufus connection, and a practical update playbook for your listings.
What is Amazon COSMO, and why does it matter now?
COSMO — Common Sense Knowledge for E-commerce — is the AI framework Amazon began integrating into search and product discovery across 2024 and 2025. It was built to solve a specific problem: traditional keyword matching can't understand what shoppers actually mean.
A shopper searching "safe for my toddler" on legacy A9 would only surface products with those literal words in the title or backend keywords. COSMO evaluates a commonsense knowledge graph — it understands that "BPA-free," "non-toxic," "ages 1–3," and "pediatrician recommended" are all related to toddler safety, even if the search query uses none of those terms. It surfaces products based on inferred intent, not string matching.
For CPG brands, this shift changes what a "well-optimized listing" actually looks like. The brand that described its product most completely — across use cases, attributes, consumer scenarios, and category context — wins discovery. The brand that keyword-stuffed its title with every search-term variation gets deprioritized by the same system that used to reward it.
This isn't theoretical. Brands optimizing for semantic completeness — not keyword density — are seeing 20–40% share-of-voice lifts in early 2026 data. Not because they changed their ad budgets. Because they changed how their listings communicate.
Is A9 actually dead?
Not technically. A9's core ranking signals — sales velocity, conversion rate, pricing competitiveness, availability — still carry weight. Amazon didn't flip a switch and delete the old system. What Amazon did was layer COSMO on top of A9, making semantic understanding a significant weighting factor in ranking decisions alongside the transactional signals A9 always used.
Think of it as a two-layer system. A9 is still evaluating whether your product sells, converts, and is priced competitively. COSMO is evaluating whether your product should be in the consideration set in the first place — based on what the listing actually says about the product and whether it matches what the shopper actually means.
The practical result: a product with strong A9 signals (high velocity, good conversion, competitive price) but poor semantic completeness will lose share of voice to a product with moderate A9 signals and excellent semantic completeness. That's the shift most agencies haven't internalized. They're still running A/B tests on title keyword order while the ranking model has moved underneath them.
How does Rufus fit into the picture?
Rufus is Amazon's customer-facing AI shopping assistant. COSMO is the intelligence layer behind it. When a shopper asks Rufus "what's the best protein powder for runners who hate chalky taste," COSMO processes that query through its knowledge graph and returns products whose listings demonstrate relevance to running, protein supplementation, taste quality, and smooth texture — regardless of whether the listing literally contains the word "chalky."
Rufus crossed 40M+ monthly active shoppers by Q2 2026. That's not a beta anymore. That's a primary discovery channel running alongside traditional search. And Amazon has confirmed that Rufus-influenced purchases show up in Brand Analytics — meaning brands can see when Rufus is driving traffic and conversions to their listings, or when it isn't.
The problem this creates for brand managers is clear: if your listing doesn't answer the kinds of questions Rufus fields — "is this safe for kids," "does this work for sensitive skin," "what's the protein per serving," "will my dog actually eat this" — you're invisible to a growing share of Amazon shoppers who aren't typing keywords. They're asking questions. And COSMO is answering those questions with your competitor's product.
What are the four signals COSMO reads?
Here's a framework we use at Neato to audit every listing for COSMO readiness. We call it The Four Signals COSMO Reads.
Signal 1: Semantic completeness of listing text. COSMO evaluates how fully your listing describes your product across attributes, use cases, consumer scenarios, and category context. A pet supplement listing that only describes ingredients is semantically incomplete. One that also addresses the pet's age range, the health conditions it supports, how it's administered, its taste profile, and what a vet would consider is semantically rich. COSMO rewards the second listing because it can match it to more shopper intents across more query types.
Signal 2: Review sentiment alignment. COSMO cross-references your listing claims against what customers actually say in reviews. If your listing says "great taste" but 30% of your reviews mention chalky or gritty texture, COSMO sees a misalignment. This isn't about star ratings — it's about whether the product description matches the real-world customer experience as expressed in review language. Brands with high review-listing alignment get stronger semantic trust signals.
Signal 3: Category consistency. COSMO evaluates whether your product's attributes are consistent with the category it's placed in. A product listed in "vitamins and supplements" that describes itself primarily as a beauty product creates a category signal mismatch. This sounds basic, but brands with broad product lines frequently have ASINs miscategorized or straddling two categories with listing copy that doesn't fully commit to either one.
Signal 4: Cross-product entity connections. This is the knowledge graph layer. COSMO maps relationships between products, ingredients, use cases, and consumer needs across its entire product graph. A collagen supplement that mentions joint health, skin elasticity, post-workout recovery, and bone density support is connected to more entity nodes than one that only describes collagen content. The more entity connections your listing activates, the more search contexts — and Rufus query types — COSMO can place your product into.
How should CPG brands update their listings for AI discovery?
This isn't a full content overhaul. It's a targeted set of updates to how your listings communicate with an AI system that reads for meaning, not keywords.
Rewrite bullet points for questions, not keywords. Instead of "organic grain-free dog treats made with real chicken," write copy that answers implied questions: "Made with USDA organic chicken as the first ingredient. Grain-free formula designed for dogs with wheat or corn sensitivities. Sized for small to medium breeds, 5–40 lbs." The first version targets keyword matches. The second answers the questions a shopper — or Rufus — would actually ask.
Add use-case scenarios to A+ content. A+ is prime COSMO territory. Describe when, why, and how consumers use the product in real life. "Morning routine for joint stiffness" is more semantically useful than "daily supplement." "Post-gym protein shake that mixes in water without a blender" is more useful than "easy-to-mix protein powder." Specificity is what COSMO rewards.
Audit your review-listing gap. Read your top 50 reviews. Identify the three attributes customers mention most that your listing doesn't. If customers praise something your listing never references — say, "my dog actually eats this one" for a pet supplement — COSMO can't connect the dots between that shopper query and your product. Close the gap.
Fix category placement. Check your Browse Node assignments and product type keywords. If an ASIN is miscategorized, COSMO's category consistency signal is working against you. This is a 15-minute fix that most brands never perform because nobody thinks to audit it.
Expand backend search terms to semantic neighbors. Don't just add keyword synonyms. Add related concepts, use cases, and consumer needs that connect to your product's entity graph. A joint supplement for dogs should have backend terms like "hip mobility," "senior dog health," "glucosamine chondroitin," and "post-surgery recovery support" — concepts that expand the product's semantic footprint, not just keyword variations of the main term.
What's the Neato point of view on COSMO and AI discovery?
We spend more time on listing optimization than any other single operational activity at Neato — more than advertising, more than inventory forecasting. That was true before COSMO, and it's more true now. Because in a COSMO-driven discovery environment, the listing isn't just a product page. It's the input that an AI system reads to decide whether your product exists for a given shopper's query.
Neato operates as a 2P eCommerce accelerator, which means we own the content strategy and listing execution for every brand we partner with. When COSMO shifts the ranking rules, we update every listing across the portfolio — not next quarter, but within weeks. That operational speed is a structural advantage because the brands that adapt first capture the share of voice that slower competitors lose, and share of voice in a COSMO world compounds faster than it did in an A9 world.
The brands still paying an agency to A/B test titles with different keyword orderings are optimizing for a system that no longer drives the results they're measuring. COSMO rewards completeness, accuracy, and consumer empathy in listing copy. Those aren't SEO skills. They're brand management skills. And they're why brand managers — not just Amazon keyword specialists — need to be leading listing strategy in 2026.
