How ChatGPT picks the products it shows — and how a small brand gets in

⏰ The 30-second version

• ChatGPT shopping results have two layers: a product carousel fed by structured product data, and the written recommendation around it, which is assembled from ordinary web pages.
• Semrush ran the same product query 100 times and found ChatGPT's top product appeared in Google Shopping's first three results 75% of the time. Your Google Shopping health leaks straight into the carousel.
• Shopify merchants are already in the catalog integration by default. Non-Shopify merchants apply for a direct feed at chatgpt.com/merchants.
• The feed gets you into the carousel. It does not write the sentence that says why you're the right choice — that comes from content, reviews and third-party pages.
• Small brands lose in the written layer more often than in the feed. That's also the layer you can fix this week without an integration.

Two different machines

When someone asks ChatGPT "what's a good fragrance-free sunscreen for sensitive skin", two things happen at once, and they run on different fuel.

The carousel. Products with images and prices. This is powered by structured product data — a feed pushed to OpenAI, plus third-party product metadata. OpenAI's help documentation says merchants are ranked on factors like availability, price, quality and whether they're the maker or primary seller, and that results are organic rather than paid placement.

The answer. The paragraphs explaining what matters for sensitive skin, which ingredients to avoid, and which brands people rate. This is retrieved from the open web: guides, reviews, forums, and your own pages if they happen to answer the question in extractable text.

Most DTC brands obsess over the first machine and ignore the second. The second is where "and X is the one usually recommended for this" gets decided.

What the evidence says about the carousel

Semrush's experiment is the most useful public data point. Running a product query 100 times, they found ChatGPT issues two sets of background searches: contextual queries that build the written answer, and Google Shopping queries that populate the products. The top ChatGPT product sat in Google Shopping's top three results 75% of the time.

The practical read: for the carousel, your Google Merchant Center feed quality, pricing, availability and review volume are doing more work than anything you write.

Meanwhile the official path keeps getting more formal:

• Shopify merchants are integrated through Shopify Catalog. OpenAI states no additional work is required from individual merchants.
• Everyone else applies at chatgpt.com/merchants and, once approved, pushes a feed (CSV, TSV, XML or JSON) to an allow-listed HTTPS endpoint, refreshable as often as every 15 minutes.
• Feed fields include explicit flags — is_eligible_search controls whether a product can appear in ChatGPT search results at all.
• Shopping is live for ChatGPT users in the U.S. first, expanding from there.

If you sell on Shopify in the U.S., you are probably already eligible and don't know it. Check by asking ChatGPT a category question and seeing whether you appear.

Where small brands actually lose

We scan a lot of small ecommerce sites. The pattern repeats:

The answer layer never had a chance. The brand's product page describes the product in marketing language and puts the useful detail — ingredients, who it suits, what it doesn't do — in an image. A model extracting text finds a slogan and moves on.

No comparison content exists. Buyers ask comparative questions ("X vs Y for oily skin"). The only pages answering those are affiliate roundups, so the affiliate roundup decides who gets named.

Third-party evidence is thin. Reviews concentrated on your own site help conversion but travel poorly. Mentions on independent sites, forums and retailer pages are what the answer layer reads.

The buyer question was never written down. Teams optimise for the phrase they use internally, not the sentence a customer types. "Barrier-repair moisturiser" versus "moisturizer for oily but dehydrated skin" — one of those gets asked.

An order of work that fits a small team

1. Feed hygiene first, because it's mechanical. Google Merchant Center clean, titles descriptive, availability accurate, Product schema on product pages. If you're not on Shopify, apply for the OpenAI feed.
2. Move the buying detail out of images into text. Ingredients, sizing, who it's for, who it isn't for. Same page, readable format.
3. Write the comparison you're currently outsourcing. One honest page comparing your product to the obvious alternative, including where the alternative wins. Models quote balanced pages more readily than one-sided ones.
4. Answer the top five questions your support inbox already gets. In the customer's words, in an FAQ block, conclusion first.
5. Re-ask the same questions 30 days later. Same questions, same models. Otherwise you're measuring noise.

What about Perplexity and the others

Perplexity's shopping surface leans on its own retrieval plus merchant data, and Google's AI Mode leans on the Shopping Graph it already owns. The mechanics differ, but the two-layer shape holds everywhere: a structured product layer you feed, and a written recommendation layer assembled from pages anyone can publish.

That second layer is the one a small brand can win, because it's decided by whoever answered the question best, not by whoever has the biggest catalog.

The free scan shows which of 30 real buyer questions your page currently answers, and what people actually search around them. It takes about 30 seconds and doesn't ask for a card. For the full picture of what an audit covers, see what a GEO audit measures.

FAQ

Q. How do I get my products into ChatGPT shopping results? If you're on Shopify, product data already flows through Shopify Catalog. Otherwise apply at chatgpt.com/merchants for direct feed access, then push a compliant feed with is_eligible_search set to true.

Q. Can I pay to appear in ChatGPT shopping results? OpenAI states shopping results are organic and not influenced by paid placement. Feed quality, relevance and user context determine what surfaces.

Q. Does ChatGPT rank products by price? Not directly. OpenAI's documentation says merchants are ranked on factors including availability, price, quality and whether they are the maker or primary seller, and price shown may not be the lowest available.

Q. Why does ChatGPT recommend my competitor and not me? Usually because the written layer, not the product layer, named them: a review site, a comparison page or a forum thread answered the buyer's question and cited them. That's content you can compete with.