How does AI shopping work for ecommerce brands?
A shopper describes what they want to ChatGPT or Gemini and gets back an answer with product cards: an image, a price, and a way to buy on the spot. That answer is the new shelf.
If your product is in the answer, you are on the shelf. If the buy link points somewhere you do not own, someone else takes the sale. This article explains how the system works end to end, and why 63% of brand recommendations never become a clean path to the brand’s own store.
The buying decision moved from the search page into the chat.
Shopping used to start on Google or Amazon. Now, more and more, it starts in a chat box.
“I’m shopping for a nice weighted blanket for a hot sleeper. Show me some good options.”
Instead of ten blue links, the shopper gets an answer: a product, a price, an image, and a way to buy on the spot. That answer is really the product itself, surfaced and ready to buy. It has no homepage and no logo, but it is the new shelf, and it is where the buying decision now gets made.
Which AI assistants matter for ecommerce right now?
Two. ChatGPT and Gemini own the crowd, and nothing else is close.
Those are global numbers. Run deliberately conservative napkin math: assume only 30% of that usage is in the US (about 1.65 billion for ChatGPT, about 800 million for Gemini), and assume just 1 in 10 of those users touches shopping. That still leaves 200 million-plus shopping moments a month across the two platforms — with lowball assumptions at every step, a real shopping share almost certainly higher than 1 in 10, and growth every month.
Sit with that number. 200 million shopping moments a month is not a side channel. It is a primary storefront.
And not every assistant has this. The real divide is not who can mention a product — it is who has the shopping carousel: the product-card UI with images, prices, and buy links.
Almost every assistant can recommend a product in text. Only ChatGPT and Gemini turn that recommendation into a real shelf, with cards, images, prices, and a path to buy. For now, the entire game is those two.
What does an AI shopping query look like?
The query is just the shopper’s sentence. Not keywords — a full natural request, the way they would ask a knowledgeable friend in a store:
“I want a weighted blanket that won’t make me overheat, under $200, good reviews. What should I get?”
Packed into one sentence: a product type, a constraint (won’t overheat), a budget (under $200), and a quality bar (good reviews). A human salesperson juggles all four at once. So does the assistant.
What comes back is the shelf: a product card (image, name, price, sometimes a rating), a line of reasoning (“breathable and machine-washable, good for a hot sleeper”), and a link or buy button.
Two quirks that make AI shopping different from search.
Neither of these exists in normal search, and both change how visibility has to be measured.
The same query can give different answers. Run “best weighted blanket for a hot sleeper” five times and you may get five slightly different lists — in three of them you appear, in two you vanish. A single check tells you almost nothing.
Visibility is a rate, not a yes or no.
The assistant does not just read the query, it reads the person: their history, their lean toward luxury or budget, even color preferences. The same request can surface completely different brands for different shoppers.
You win by being unmistakably clear about who you are and who your ideal customer is, so the assistant can confidently match you to the right buyer.
PDP — Product Detail Page: the exact “buy it now” page for one specific item in one specific size and color.
Not your homepage. Not a category page. Every product on your store has one.
The buy button in an AI result is supposed to point at a PDP. The only question that matters is whose. Yours? A retailer’s? A marketplace listing? Or, surprisingly often, nothing usable at all?
The three kinds of results ChatGPT returns.
ChatGPT runs the most developed shopping system, so it is the clearest place to see how a query plays out. Gemini behaves similarly.
Mode 01 · product cards
Cards with real buy paths.
The strongest outcome. Product cards that link out to a store — sometimes several stores per product — down to the direct product page. A shelf with a checkout attached.
Mode 02 · info panel
Authoritative, but no buy path.
A clean summary about a brand or product, with no card and no link to buy. Looks great, sells nothing.
Mode 03 · plain text
A mention in a sentence.
Your brand named in passing. No card, no image, no link. The shopper is left to go find you on their own.
Only the first is a real shelf. The other two name you and then strand the shopper. Landing in those product cards, with a clean link to your own store, is the entire game.
Why AI shoppers convert 4–5× higher.
The brand that lands in those product cards does not just get a visit, it gets a pre-sold visit. By the time the assistant points someone at a store, it has already compared the options and made the case. The shopper arrives ready to buy.
The rails underneath: ACP, UCP, and AP2.
Buying inside the chat runs on new commerce protocols. You do not build them. Just know they exist, because they explain why this is permanent.
The tell: Shopify and Stripe, the rails most online stores already run on, wired straight into both assistants. Over a million Shopify merchants are now shoppable inside ChatGPT. When the checkout rails get laid into two platforms, those two become the storefront.
The problem: showing up is not the same as winning.
You can be the most-recommended brand in your category and still lose every sale. Every AI recommendation lands as one of three outcomes.
Outcome · winner
Winner
The assistant recommends you and the buy path lands on your own product page. Owned demand.
Outcome · leak
Leak
It recommends you, but the buy path goes to a retailer, reseller, or marketplace. You earned the credit, someone else gets the sale.
Outcome · loser
Loser
It mentions or shows you, but there is no usable buy path at all. A recommendation that goes nowhere.
We measured this across hundreds of real ChatGPT and Gemini shopping conversations in 11 ecommerce categories — roughly 380 brand observations in total. The result:
The two platforms fail in opposite ways. ChatGPT often names a product but gives no way to buy it. Gemini attaches far more buy buttons, but a large share leak to third-party sellers. The brands showed up. They still lost the transaction.
How to check your own store.
AI visibility without buy-path control is not owned demand. And you cannot fix what you cannot see.
Remember the two quirks: results are not deterministic and they are personalized. So you do not test once. You test a spread of real shopping queries, several times each, and watch how often you land as a Winner, a Leak, or a Loser. That rate — not a single screenshot — is your actual AI shelf position.
Where is ChatGPT sending your buyers?
We run real shopping prompts for your category and show you where the recommendations land — your store, a reseller, or nowhere. One form, no meeting.
Common questions.
What is AI shopping?
AI shopping is when a buying decision happens inside an AI assistant instead of a search engine. A shopper describes what they want to ChatGPT or Gemini in a normal sentence, and the assistant answers with specific products: an image, a price, a short line of reasoning, and a link or button to buy.
What is AEO (Answer Engine Optimization) for ecommerce?
AEO is the practice of making a brand and its products legible to AI assistants so they appear in AI-generated answers. For ecommerce, that means being recommended inside ChatGPT and Gemini shopping results with a working buy path to the brand’s own product page — not just being mentioned in text.
Which AI assistants have a real shopping interface?
ChatGPT and Gemini both render a product-card shopping interface with images, prices, and buy links. Perplexity has a shopping UI but far fewer users. Grok, DeepSeek, and Claude can only mention products in plain text, so for ecommerce the game today is ChatGPT and Gemini.
What is a PDP in ecommerce?
A PDP, or Product Detail Page, is the exact buy-it-now page for one specific item in one specific size and color — not the homepage and not a category page. The buy button inside an AI shopping answer is supposed to land on a PDP; the only question is whether that PDP belongs to the brand or to someone else.
How often do AI shopping recommendations send buyers to the brand’s own store?
In Caeliai’s GENESIS-03 field study of real ChatGPT and Gemini shopping conversations across 11 ecommerce categories, only about 37% of brand recommendations landed on the brand’s own product page. Around 18% leaked to a retailer or marketplace, and about 45% had no usable buy path at all — so 63% of recommendations did not become a clean path to the brand’s own store.
AI assistant usage from SimilarWeb (2026, global estimates). Conversion figures from Semrush and Adobe (2025–26). Buy-path data from Caeliai GENESIS-03 (PDF), a field study of real ChatGPT and Gemini shopping conversations across 11 ecommerce categories.