"Does the AI know about my product" is really four separate, differently-documented questions, because Rufus/Alexa for Shopping, ChatGPT Shopping, Perplexity, and Google's AI Mode each pull from a different mix of sources and disclose that mix to very different degrees. None of the four platforms publishes a complete, technical account of its ranking logic — treat everything below as accurate to what's actually been said publicly, with the gaps marked as gaps rather than filled in with guesswork.
The short version
| Platform | Confirmed data sources | How products get in at all | What's undocumented |
|---|---|---|---|
| Amazon Rufus / Alexa for Shopping | Amazon's product catalog, customer reviews, community Q&A, shopping/browse history, "information from around the web" | Automatic — pulls from your existing Amazon listing, no separate opt-in | Exact weighting between catalog attributes, review sentiment, and web content; whether backend search terms are used the same way as in organic search |
| ChatGPT Shopping | Structured metadata from first- and third-party providers (price, description, availability), reviews, user Memory/Custom Instructions | Automatic for Shopify merchants (via Shopify Catalog); everyone else applies for direct product-feed access | Full ranking formula behind which products appear in a carousel and in what order |
| Perplexity ("Shop like a Pro" / "Buy with Pro") | Merchant-submitted product data for enrolled merchants, plus Perplexity's general web index/citations for open queries | Free sign-up to the Perplexity Merchant Program; enrollment is described as improving "chances of being a recommended product," not guaranteeing it | Ranking logic for the underlying recommendation model; how heavily non-enrolled, organically-indexed products can still surface |
Amazon: Rufus and Alexa for Shopping
Amazon's own 2024 description of Rufus (quoted from an Amazon spokesperson) was that it is "a generative AI experience that's been trained on the product catalog, customer reviews, community Q&As, and information from around the web," built on an internal LLM Amazon trained specifically for shopping. That statement is the closest thing to an official mechanism disclosure this space has, and it's now three-plus years old — Amazon has not published a more detailed technical breakdown since.
In May 2026, Amazon merged Rufus with Alexa+ (its subscription voice assistant) into a unified Alexa for Shopping, which Amazon describes as drawing on: Amazon's catalog and "product expertise," personal shopping history and "conversations from across both Amazon.com and Alexa" (meaning voice and app-chat interactions now feed a shared profile), and the ability to "shop from other online stores across the web" and pull "in-depth information from across the web" for comparisons outside Amazon's own catalog. Amazon Advertising's own guidance to brands (published separately from the consumer-facing product announcement) tells sellers that "inconsistent product information, sparse descriptions, or missing context" are reasons the assistant will skip a product, and specifically names complete Brand Registry detail, full A+ content, and substantive customer reviews as the inputs that give the assistant "the fuel AI needs to draw the right information."
What's confirmed: catalog attributes, reviews, and Q&A are inputs. What's not confirmed by Amazon: any specific weighting formula, whether backend (invisible) search terms feed the assistant the same way they feed traditional A9/A10 search, or how the "information from around the web" component is selected and trusted. Third-party seller-tool blogs (Perpetua, ZonGuru, Jarvio, evolveAMZ) publish detailed optimization checklists for Rufus; where this resource repeats a specific claim from one of them, it's flagged as such rather than presented as Amazon's word.
ChatGPT Shopping
OpenAI's own help documentation describes ChatGPT's shopping results as drawing on "structured metadata from first-party and third-party providers (e.g., price, product description)," with ranking inside a product carousel influenced by "availability, reviews, pricing, and user preferences" plus the user's Memory and Custom Instructions. This is meaningfully more transparent than Amazon's disclosure, but it still stops short of a real ranking formula.
Getting a product into that pool works one of two ways: merchants selling on Shopify get automatic inclusion through a direct integration OpenAI calls Shopify Catalog, and everyone else has to apply for direct product-feed access through OpenAI's developer documentation and submit a structured feed. Separately, OpenAI's Instant Checkout, built on the open-sourced Agentic Commerce Protocol (ACP), lets a shopper complete a purchase inside the chat for participating merchants — OpenAI has said merchants pay a small fee on completed purchases and the feature is free to shoppers. At launch this covered U.S. Etsy sellers, with Shopify merchants including Glossier, SKIMS, Spanx, and Vuori announced as "coming soon" — by the time you're reading this, check OpenAI's own documentation for the current merchant list, since CNBC reported in March 2026 that OpenAI was already revising the shopping experience after early friction with the original Instant Checkout rollout. This is a fast-moving product, not a settled one.
Perplexity ("Shop like a Pro" / "Buy with Pro")
Perplexity's own description of its shopping feature emphasizes that product cards are "unbiased recommendations, tailored to your search by our AI" rather than paid placements, sourced from "live details on all the best available products" via Shopify and other participating merchants. The free Perplexity Merchant Program is the enrollment path: joining puts a merchant's catalog into Perplexity's product index and is described as increasing the "chances of being a recommended product" — deliberately not phrased as a guarantee. Buy with Pro lets Perplexity Pro subscribers in the U.S. complete checkout for eligible products without leaving Perplexity, with a redirect to the merchant's own site when that's not available.
What's genuinely unclear from Perplexity's own materials: how much weight non-enrolled products still get purely from Perplexity's general web index and citation behavior (Perplexity is fundamentally a web-search-and-synthesis product, so a well-optimized, well-reviewed product page can plausibly still get cited even without merchant enrollment) versus how much the enrolled, structured-feed path dominates actual shopping-card recommendations. Third-party guides (Stellagent, Alhena, 1Digital) describe specific optimization tactics for Perplexity Shopping; none of them are confirmed by a Perplexity technical document, so treat that guidance as directional.
The pattern underneath all four
Every one of these assistants combines two fundamentally different kinds of input: structured data the platform trusts because a merchant or the platform itself submitted it directly (a Shopify Catalog sync, a Merchant Center feed, Amazon's own catalog record), and unstructured content the assistant retrieves and synthesizes from the open web or from reviews/Q&A (closer to how a search engine or a retrieval-augmented LLM works). Optimizing for AI shopping assistants means treating both channels as real and separately manageable — see Optimizing Listings for AI Shopping Assistants for the practical version of that split, and Structured Data and Feeds for AI Discovery for the feed/markup mechanics specifically.
Common mistakes
- Assuming one platform's confirmed mechanics (Amazon's reviews-and-Q&A input, for instance) apply the same way to a different platform without checking that platform's own documentation.
- Treating a third-party optimization blog's specific tactical claim ("Rufus weights backend search terms at X%") as a documented fact rather than an educated guess — none of the four platforms has published that level of detail.
- Ignoring the open-web-retrieval half of these systems entirely and optimizing only the structured feed, when unstructured content (reviews, Q&A, even third-party articles about your product) demonstrably feeds at least Amazon's and Perplexity's assistants.
- Assuming enrollment in a merchant program (Perplexity Merchant Program, ChatGPT direct feed access) guarantees surfacing — every platform's own language is deliberately non-guaranteeing.
Best practices
- Read each platform's own documentation (OpenAI's Help Center, Amazon Advertising's brand guidance, Perplexity's Merchant Program page) directly at least once, since third-party recap articles sometimes overstate certainty the source material doesn't have.
- Treat structured-feed submission and unstructured content quality (reviews, Q&A, product-page copy) as two separate workstreams, not one.
- Re-check this landscape at least quarterly — every platform covered here changed a material mechanic (a rebrand, a checkout protocol, a merchant program) within the twelve months before this was written.
- When a claim in an outside guide isn't attributed to the platform itself, treat it as a hypothesis worth testing on your own listings rather than a rule to follow blindly.
FAQ
Is Rufus still a separate product from Alexa+? As of Amazon's May 2026 announcement, no — the two were merged into a single "Alexa for Shopping" experience spanning the Amazon app and Alexa/Echo devices. Older content (including some cited in this cluster) still refers to "Rufus" by name because that was the accurate, separate product name before the merger.
Do these assistants use my Amazon/Google/Shopify SEO the same way traditional search does? Partially, and unconfirmed in detail. All four platforms draw on some of the same underlying structured data (titles, attributes, feeds) that traditional search and marketplace search already use, but none has confirmed that ranking signals transfer one-to-one — see GEO/AEO vs. Traditional SEO, Explained for why that distinction matters.
Which platform is most transparent about its mechanics? Of the four, OpenAI's help documentation for ChatGPT Shopping is the most specific about what factors influence ranking (availability, reviews, pricing, user preference) and how to get included (Shopify Catalog or direct feed application). Amazon and Perplexity have published less operational detail; both rely more heavily on third-party interpretation to fill the gap.