36:33
Amazon Rufus
Amazon's new Cosmo algorithm now powers Rufus, its conversational AI shopping assistant, as a system distinct from the legacy keyword-focused A9 algorithm, and sellers who don't restructure listings around Cosmo's knowledge-graph mapping — rather than pure keyword stuffing — risk losing access to a projected $10 billion in incremental Rufus-driven sales in 2026.
Cosmo is Amazon's new algorithm powering Rufus's conversational search, layered on top of / distinct from the traditional keyword-intent A9 algorithm.
Rufus can proactively recommend, add items to cart, or even purchase on a customer's behalf, without the customer using a traditional search box.
Amazon states customers who engage with Rufus are 60% more likely to purchase, with 250 million customers using Rufus in 2025 and a projected $10 billion in additional 2026 sales tied to it.
Amazon's urgency is framed as competitive: ChatGPT has partnered with Shopify and Walmart, and Amazon wants to keep the research-to-purchase flow inside Rufus rather than lose it to ChatGPT or Perplexity.
Cosmo dynamically routes queries to different underlying LLMs (via Amazon Bedrock, including Amazon Nova and Anthropic models) depending on whether a query is a simple lookup or research-heavy.
Cosmo/Rufus draws on five data sources: catalog data, reviews, Q&A, external web/publication sources, and customer behavior.
Amazon reportedly does not pull from Reddit or Quora for external sources, instead partnering with specific curated review/publication sites that Rufus cites directly.
Rufus incorporates RAG (Retrieval-Augmented Generation), layering live external-data retrieval on top of its base LLM knowledge, which the video credits with lifting Rufus's quality to be comparable to ChatGPT.
Listings need a 'knowledge graph'/technical mapping of attribute nodes (e.g., 'used with,' 'used in location') so products surface across many conversational use cases beyond literal keyword matches.
Keyword-stuffing is framed as an outdated signal; sellers are now freer to write brand-story-driven copy for human conversion, provided separate technical attribute mapping handles algorithmic matching.
External activity off Amazon — TikTok Shop, Shopify traffic, promotions, and PR/publication placements — is said to spill over into on-Amazon visibility and Rufus trust signals.
ZonGuru offers a free 'Cosmo Readiness Report' (scored out of 100) diagnosing mapping and Q&A coverage, and a paid 'Cosmo Transformation Service' (multi-agentic AI plus human experts) that rewrites listing content, images, and EBC content.
ZonGuru claims listings moved from keyword-stuffing to intelligent mapping see sessions and conversion rate increase, while listings left unchanged for about 6 months are reportedly falling in rank.
Cosmo — Cosmo is Amazon's new algorithm that powers Rufus's conversational search, distinct from the keyword-intent-focused A9 algorithm, dynamically routing queries to different LLMs and drawing on catalog, review, Q&A, external-web, and customer-behavior data. Apply: Map listing content and attributes to Cosmo's data sources via a knowledge graph rather than relying only on keyword density, since Cosmo now governs whether a product surfaces in Rufus's conversational recommendations.
A9 — A9 is Amazon's traditional, keyword-intent-focused search algorithm that Cosmo is described as evolving beyond for conversational search. Apply: Keep optimizing for A9-style keyword search, since customers still type keywords and convert on listing copy, while layering Cosmo-specific mapping on top rather than replacing one approach with the other.
Rufus — Rufus is Amazon's AI shopping assistant/conversational search agent that can answer questions, recommend products, add items to cart, or even purchase on a customer's behalf. Apply: Test your own listings by typing customer-style questions into Rufus to see whether your product surfaces, then adjust content where it doesn't.
RAG (Retrieval-Augmented Generation) — RAG combines a trained LLM with live retrieval of external data layered on top of its base knowledge to produce a better answer; the video says most LLMs use it and Amazon added it to Rufus. Apply: Recognize that Rufus pulls live external signals (reviews, curated publications, customer behavior) alongside listing content, so strengthening those external signals feeds directly into what Rufus retrieves and recommends.
Knowledge graph / technical mapping — A knowledge graph is the technical mapping of a listing's attributes into nodes that connect the product to many real-world use cases and query types, beyond literal keyword matches. Apply: Build out attribute nodes on each listing so Cosmo/Rufus can surface the product across conversational queries that never mention the product's own keywords.
"Used with" attribute node — A knowledge-graph attribute mapping a product to items it is commonly used alongside, e.g., an espresso maker mapped to coffee beans, coffee mugs, and morning/breakfast use. Apply: Add explicit 'used with' associations to listing attributes so that when a customer asks Rufus about a related item, the mapped product gets recommended as a co-buy.
"Used in location" attribute node — A knowledge-graph attribute mapping a product to the location or context where it is typically used, cited as commonly missing from listings. Apply: Fill in location-of-use attributes on listings to capture Rufus queries framed around where a product is used.
Multi-LLM routing (multimodal usage) — Cosmo selects which underlying LLM answers a given Rufus query depending on query type — e.g., a simple price-lookup query versus a research-heavy, expert-referencing query — drawing on models like Amazon Nova and Anthropic's models. Apply: Write listing content that supports both simple factual lookups and deeper research-style framing, since a product may need to satisfy either kind of query depending on which model Cosmo invokes.
Amazon Bedrock — Amazon Bedrock is described as Amazon's underlying AI orchestration platform into which multiple LLMs can be plugged to build AI processes such as Cosmo. Apply: Treat Bedrock as background infrastructure context for why Cosmo can flexibly route between different LLMs, not something sellers configure directly.
Cosmo Readiness Report (Cosmo RS) — A free ZonGuru diagnostic tool that takes an ASIN and returns a score out of 100 for how well the listing is technically mapped to Cosmo plus its Rufus Q&A coverage, usable up to 5 times for free. Apply: Run your ASINs through the Cosmo Readiness Report to identify mapping gaps before deciding whether to fix listings yourself or pay for the transformation service.
Cosmo Transformation Service — ZonGuru's paid, multi-agentic AI plus human-expert service that researches brand positioning, rewrites listing copy/images/EBC content, and maps attributes to the knowledge graph for Rufus visibility; a single-listing run costs $75. Apply: Use the service, starting with one $75 listing, to get a done-for-you rewrite and mapping, then measure session/conversion lift before rolling it out to the full catalog.
Listing copy optimization checklist (human + AI dual-optimization) — A set of copywriting practices recommended for listings that must convert human readers and be retrievable/citable by Rufus: write for human conversion, answer intent rather than just keywords, use short/sharp sentences, map feature→benefit→use case, optimize images/alt tags/metadata, align claims with Q&A, and pursue PR/publication placements. Apply: Rewrite listing copy, images, and Q&A against this checklist so the same content both converts human shoppers and is easy for Rufus's retrieval layer to parse and cite.
Rufus's described ability to autonomously add items to cart or complete purchases on a customer's behalf frames a shift from search-and-click toward agentic commerce.
Amazon reportedly excludes UGC forums like Reddit and Quora from Rufus's external web sources, instead sourcing from curated review/publication sites — implying that forum-seeding tactics wouldn't feed Rufus directly, only PR placement with specific publications would.
Cosmo's personalization is said to weight roughly the last 2–3 weeks of customer behavior rather than all-time history, meaning near-term behavioral signals matter more than long-run purchase patterns.
The narrative casts Rufus's growth as existential for Amazon (own the conversational-commerce space or lose market share to ChatGPT/Perplexity), positioning listing-optimization advice within a platform-competition story rather than a routine algorithm tweak.
The pitch reframes brand-storytelling copy — previously seen as a marketing nice-to-have — as now compatible with algorithmic visibility, as long as a separate technical/attribute mapping layer carries the keyword-matching burden.
Cosmo is described as selecting different underlying LLMs depending on query complexity (simple price lookup vs. expert-referencing research query), implying a single listing must satisfy both simple factual retrieval and deeper narrative framing.
«Your Amazon listing isn't written for customers anymore. It's written for the AI that advises them.»
— 05:35
«It's all about asking questions now and getting and giving the answer rather than just mapping for keywords.»
— 07:54
«If you used Rufus 6 months ago, you'd have a pretty terrible experience, but right now it is actually a very comparable experience to any of the other LLMs that are out there like ChatGPT.»
— 09:06
«Rufus is the competitor and it has to move fast because if it doesn't own that space, it's giving it away to these other LLMs.»
— 10:50
«If we're not giving the right signals to Rufus and getting recommended in that conversational search, you are absolutely losing your chance to get access to that $10 billion in additional sales.»
— 11:56
«no longer do you need to just stuff your listing with all these keywords, because that's the signal that you're giving to to Cosmo and and Rufus.»
— 20:57
«if you don't have the mapping, you're not going to appear for all of these different searches and all of that traffic across that $10 billion in additional revenue that Amazon is generating.»
— 23:21
«you've got to get in there and get it done and uh and do it now so that you can capitalize on that before your competitors do it.»
— 32:01
«in order for Amazon to give the best conversational search experience, it has to use a new intelligent algorithm that they are all in on, which is called Cosmo.»
— 33:20
«I hope this video will help you to optimize your listings for AI-driven product discovery.»
— 35:44
Reception
No comments were available to gauge audience reception.
This is a vendor-hosted interview (ZonGuru's founder) that surfaces real, specific claims about Amazon's Cosmo algorithm and Rufus's growing role, but the explanatory content is structured to funnel viewers toward ZonGuru's free report and paid Cosmo Transformation Service, so its urgency framing and statistics should be read as coming from an interested party rather than an independent analysis.
36:33