ChatGPT shopping: how products get recommended

What ChatGPT shopping is

ChatGPT shopping is what happens when someone asks ChatGPT what to buy. Instead of ten blue links, they get a short list of named products, a reason for each, and a link to the store.

People do not type keywords into ChatGPT. They describe the situation, the budget and the deal-breakers, then ask follow-up questions until they are ready to buy:

• “Best carry-on suitcase under $200 that fits Ryanair sizing”
• “Quiet robot vacuum for a small flat with a dog”
• “Compare the top three standing desks for someone 6'4"”
• “Moisturiser for sensitive skin without fragrance, under $30”

Every one of those answers names a few products and leaves the rest out. For a merchant, that is the whole game: there is no page two, and a shopper who gets a confident recommendation rarely goes looking for alternatives.

What changed, and when

ChatGPT shopping has changed shape several times in a year. These are the milestones that matter for merchants:

The direction is clear. ChatGPT has moved from trying to own checkout to owning discovery: it reads structured product data at scale, makes the recommendation, and hands the shopper to your site to pay. That makes being recommended the thing to optimise.

How products get recommended

OpenAI does not publish a ranking formula for product recommendations, so nobody can give you the exact weights. What you can see, by asking questions and reading answers, is what the recommended products tend to have in common. Four inputs do most of the work:

The question decides the shortlist

A shopper asking for a “quiet robot vacuum for a small flat with a dog” has given ChatGPT three filters. Products whose data or coverage says nothing about noise, size or pet hair are hard to match, however good they are. This is why complete attributes matter more in AI shopping than in a keyword search box.

Reviews and roundups explain the pick

ChatGPT usually gives a reason for each recommendation, and those reasons tend to echo what reviewers and roundup articles say. If the well-known “best of” lists in your category leave you out, or describe you with an attribute you do not lead on, expect the answer to do the same.

Consistency builds confidence

When the price, availability and specs in your feed, on your product page and in your schema agree, there is less for the answer to hedge on. When they conflict, the safest recommendation is often someone else.

What merchants should do

1. 1

   Get your catalog where ChatGPT can read it

   On Shopify, check that your products are available through Shopify Catalog and Agentic Storefronts. Not on Shopify? Shopify's Agentic plan is one route to listing products for AI agents.
2. 2

   Fill in the attributes shoppers ask about

   Materials, dimensions, compatibility, dietary or skin-type suitability, what is in the box. Put them in titles and attributes, not only in images or long descriptions.
3. 3

   Add Product, Offer and Review schema that validates

   Name, brand, GTIN or SKU, price, currency, availability and aggregate rating on every product page, kept in sync with your feed.
4. 4

   Answer constraint questions on the page

   Shipping regions, returns, sizing guidance and comparisons with the obvious alternatives. Specific answers are easy to quote.
5. 5

   Earn coverage in the sources ChatGPT cites

   Find the review sites and roundups that appear in answers for your category and work on getting included, accurately described.

None of this is exotic. It is the same product data hygiene and PR work that helps everywhere, pointed at the questions shoppers actually ask. Our guide to generative engine optimization covers the content side in more depth.

How to check if ChatGPT recommends your products

You can test this yourself in an afternoon. The method matters more than the tool:

• Write 15–20 questions a real shopper would ask, mixing discovery, comparison, budget and compatibility questions. Avoid your brand name.
• Ask each one in a fresh chat with search on, so earlier answers do not colour later ones.
• Record every product named, its position in the answer, how it is described, and the sources linked.
• Repeat the same questions next month. One run is a snapshot; the trend is what tells you if your changes worked.

Then run the same questions in Perplexity, Claude and Google. A product can be ChatGPT's first pick and absent from Claude, which answers from what it learned in training rather than a live search.

Beyond ChatGPT

ChatGPT is not the only assistant shoppers use. Perplexity launched Instant Buy with PayPal in November 2025, and Google said at I/O on 19 May 2026 that AI Mode has passed one billion monthly users. Each engine uses different sources, so a brand that only watches ChatGPT sees only part of the picture.

Google's AI answers deserve special attention because they sit on top of the search results you already rely on. Our AI Overview tracking page explains how to monitor them, and answer engine optimization covers how to become the answer across engines.