Why ranking #1 in Google Shopping doesn't get you into ChatGPT's or Gemini's answers

A home goods brand’s best-selling throw blanket sits in the top three results for its exact category term in Google Shopping, has been there for months, and has the reviews and the Merchant Center feed to show for it. Ask ChatGPT “what’s a good cotton throw blanket for a couch” and the product never comes up. Ask Gemini the same question and it shows up once, then disappears the next week for a near-identical phrasing. Nothing about the product changed. The team’s instinct is to assume something broke, because the product clearly works for Google Shopping and AI answers are supposed to be reading the same web.

They aren’t reading the same thing, and for one of the two engines they aren’t reading anything connected to Google Shopping at all. That’s a specific, checkable fact about how each pipeline is built, not a mysterious drop in quality.

ChatGPT’s shopping results run on a feed Google never sees

OpenAI publishes its own product feed specification for ChatGPT shopping, separate from anything a merchant submits to Google. The spec defines nine required fields per item (item_id, title, description, url, brand, seller_name, image_url, availability, price), plus flags that control whether a product is eligible to be shown in search results at all, independent of whether it is eligible for ads or checkout (OpenAI developer docs, commerce specs). Nowhere in that specification does OpenAI reference Google Shopping, Google Merchant Center, or a merchant’s search ranking. The two systems are not the same graph read two ways. They are two separate submissions, and a product ranking well in one tells the other nothing, because the other has never seen it.

This is why a product can be a genuine Google Shopping winner and a complete non-entity in ChatGPT: if the merchant (or the merchant’s platform) never submitted that feed to OpenAI, or submitted it with gaps, the product simply isn’t in ChatGPT’s index to begin with. Shopify stores are an exception, since Shopify’s catalog integration feeds OpenAI automatically; most other platforms require an explicit, separate submission that has nothing to do with the merchant’s SEO or Shopping performance.

Gemini is closer to Google Shopping, but still a separate measurement

Gemini’s shopping answers are built on Google’s Shopping Graph, the same underlying product dataset that powers Google Shopping itself, which Google describes as giving Gemini shoppers “unmatched selection” from the same graph Search uses (Google, on Gemini’s shopping features). That shared foundation is the real difference from ChatGPT: a product that’s properly fed into Merchant Center has at least entered the pool Gemini draws from.

Being in the pool is not the same as being chosen from it, though, and Google treats the two as different enough to measure separately. Merchant Center’s AI performance insights, which Google rolled out through 2026, track share of voice specifically for AI Mode, AI Overviews and the Gemini app, as their own metric distinct from standard Shopping performance, and recommend filling attribute gaps and optimizing feeds without stating how AI surface selection relates to Shopping rank (Google: AI performance insights announcement). If Google itself ships a dedicated report to tell merchants how they’re doing in AI surfaces, that is itself evidence the company doesn’t consider Shopping rank and AI-surface presence interchangeable. A product can be well-ranked in Shopping and still score low on AI share of voice, and the only way to know is to check that report directly rather than infer it from the Shopping numbers.

Our guide to tracking visibility across multiple product lines covers setting up measurement that catches a split like this one instead of assuming one score represents every surface a product could appear on.

The feed dependency on the ChatGPT side is getting stronger, not weaker

This gap matters more every month because ChatGPT’s shopping answers are leaning harder on submitted feeds rather than general web retrieval. Profound’s tracking of ChatGPT Shopping responses found that the share of product picks sourced from merchant feeds, rather than the open web, jumped from 8.26% to 61.54% in a single week in July 2026, and held near 65% through September, based on a sample of roughly 1.76 million tracked prompt runs (Profound’s research, reported by Relevant Audience). The same tracking found the number of unique merchants referenced fell by more than 20% over that period, while the top 10 stores’ share of references nearly doubled. That pattern, feed reliance up and the number of visible merchants down, is what you’d expect if ChatGPT shopping answers are increasingly choosing from whoever actually submitted a usable feed rather than whoever has the best product or the best Google Shopping rank.

For a brand with a genuinely strong product that’s absent from this growing feed-first pool, the lesson isn’t to improve the product page. It’s to check whether the feed submission exists at all, since a shrinking visible-merchant pool on a feed-dependent surface is exactly the setup where a good product goes quiet for a reason that has nothing to do with quality.

Most catalogs aren’t actually built for this yet

The reason this gap is common rather than rare is that few merchants have the two systems both sorted at once. PayPal’s first U.S. Agentic Commerce Pulse Survey, polling 498 merchant decision-makers, found that nearly 95% of merchants can already observe traffic arriving from AI shopping agents, but only about one in five have structured their product catalog into a machine-readable format those agents can actually act on (PayPal’s Agentic Commerce Pulse Survey, via The Outpost). Most teams know AI traffic is happening to their site. Far fewer have done the separate, unglamorous work of making sure each AI surface’s own feed requirements are actually met, on top of whatever’s already working for Google Shopping.

Why an AI shopping assistant picks one of your products over another covers the mechanism once a product is already in an engine’s index and competing for a specific query. This is the layer underneath that: whether the product is in the index in the first place, which is a feed and eligibility question, not a comparison question.

What to actually check, in order

Before assuming a product needs better copy, reviews, or structured data, confirm each surface actually has the product at all. First, check whether a product feed has been submitted to OpenAI’s commerce program directly, separate from any Google feed, since that submission is what makes a product eligible to appear in ChatGPT shopping results at all (OpenAI developer docs). Second, confirm Google Shopping free listings are enabled in Merchant Center, since Gemini and AI Mode read from the same graph that free listings feed into, not from paid Shopping ads alone. Third, if a Merchant Center account has access to the AI performance insights report, read it directly rather than assuming Shopping rank predicts it; it’s reporting a different number for a reason. The engines worth tracking in 2026 is a useful next stop for prioritizing which of these surfaces matters most for a given buyer base, since the answer isn’t the same for every category.

A product that fails all three checks isn’t an AI visibility problem to diagnose with better content. It’s a plumbing problem: the pipe to that engine was never connected.

Check which surfaces actually have your product

Ask our comparison-features overview walks through how ChatGPT, Perplexity and Gemini each turned shopping into a comparison, which is the layer that matters once a product clears the eligibility bar covered here. If you haven’t checked that bar yet, a free Visibility Check against your best-ranked Google Shopping product will show you, same day, whether ChatGPT and Gemini can even see it. If the gap is real, Monitor tracking at the product level turns that one check into a trend you can act on before the next catalog update widens it.