How ChatGPT Shopping Picks Products (and What You Control)

Quick answer: OpenAI does not publish a ranking formula for ChatGPT shopping, so anyone offering you one is guessing. What OpenAI documents is this: ChatGPT shows products it judges relevant to what the shopper asked and to their context (budget, memory, custom instructions), it draws on structured product data from providers (for Shopify stores, that is Shopify Catalog), it ranks the merchants for a chosen product on availability, price, quality and whether you are the maker or primary seller, and it says product results are not ads. You cannot control the ranking. You can control whether your products are eligible, how complete and accurate your product data is, your store policies, your reviews and whether OpenAI’s crawler can read your site.

A founder asks why ChatGPT recommended a competitor’s product and not theirs. The honest answer starts with what nobody outside OpenAI knows: how the weights work. Many articles fill that gap with confident mechanics, and most of those claims do not appear in any OpenAI or Shopify page I could find.

This guide sorts the topic into three piles: what OpenAI documents, what Shopify says about how your products reach ChatGPT, and the guesses, which I leave unused. Then it turns the first two into things you can change on your store.

How I checked. On 30 October 2026 I read OpenAI’s shopping help article, its shopping research announcement, merchant page, product feed documentation, web search help article, crawler page, commerce policies and ads help article, plus Shopify’s help pages for agentic storefronts and Shopify Catalog. I state only what those pages say. Vendor blogs and trade articles were leads. This area changes quickly, so time-sensitive points are marked “as at 30 October 2026”.

Two things people call “ChatGPT shopping”

OpenAI’s pages describe two experiences, and they pick products slightly differently.

OpenAI announced shopping research on 24 November 2025. The announcement says the mode asks clarifying questions about budget, who the item is for and which features matter, then researches in the background for up-to-date price, availability, reviews, specs and images, and that results are organic. Shoppers can steer it by marking products “Not interested” or “More like this”. Because the answers, memory and feedback differ for each shopper, expect different shoppers to see different results.

What OpenAI documents about how products are picked

Intent and context come first

OpenAI’s shopping help article says a product appears in the results when ChatGPT perceives it as relevant to the shopper’s intent, and that ChatGPT considers the query and context such as Memory or Custom instructions. Its example is a shopper asking for dog costumes: ChatGPT may weigh available options, price, reviews and ease of use, and if the shopper has said they dislike clowns, it may leave out clown costumes.

The article is open about the limits. It says the model can occasionally misread intent and that not all available products will necessarily be shown. If a shopper states a budget, ChatGPT will focus more on price; if not, it may focus on other aspects.

My reading: you are competing for relevance to one shopper’s stated needs, not for a fixed list.

The inputs OpenAI names

The same article says that, when deciding which products to surface, ChatGPT considers three things:

1. Structured metadata from first-party and third-party providers (it gives price and product description as examples) and other third-party content.
2. Model responses ChatGPT generated before it considers any new search results.
3. OpenAI’s safety standards and product policies.

It adds that some factors matter more than others depending on the request. It gives no weights.

How merchants are ranked once a shopper picks a product

When a shopper selects a product, ChatGPT may show a list of merchants that offer it. OpenAI says that list is built from merchant and product metadata from third-party providers or directly from merchants, and that merchants are ranked on factors like availability, price, quality and whether they are the maker or primary seller of the item. OpenAI says it expects this to evolve, including more personalised merchant rankings.

Two details from the same article matter for a small brand:

• Price display. The price in ChatGPT’s first response typically comes from the first merchant listed and may not be the lowest. OpenAI says ChatGPT works to highlight the lowest-priced option it knows of in the product detail page and merchant rankings, and a “Best price” label may appear.
• Delays. When merchants change pricing or shipping terms, there may be a delay before the change shows.

OpenAI lists “maker or primary seller” as a merchant-ranking factor, so if you make what you sell, make that unambiguous: brand name, product names and seller should read the same on your product pages and everywhere else you are described.

Reviews, labels and descriptions are generated

OpenAI says ChatGPT may write simplified product titles and descriptions from provider information, because merchants often name the same product differently. Labels such as “Budget-friendly” or “Most popular” are generated by ChatGPT from information available to the model, which can include third-party data, and OpenAI says they are not guarantees or verified statements. Its example: “Budget-friendly” might mean reviewers often mention good value, not that the price is the lowest.

Review summaries are model-generated from reviews on public websites, and OpenAI says reviews and ratings are not verified by OpenAI.

The consequence: ChatGPT may describe your product in words that are not yours. You influence that wording by making the source material, your product page and your public reviews, clear and accurate.

Product results are not ads, and ads now exist separately

OpenAI’s shopping article says product results are selected independently by ChatGPT, are not ads, and are not influenced by any OpenAI partnerships. It says ads are separate from product results.

Ads do exist in ChatGPT as at 30 October 2026, so be exact. OpenAI’s ads help article says ads may appear for users on the Free and Go plans, run on separate systems from the chat model, are clearly labelled and do not influence ChatGPT’s answers. It calls them paid placements and says that, when several ads are eligible, it weighs a combination of factors including relevance and advertiser bids.

So “no paid placement” is accurate for product results as OpenAI describes them, and “no ads at all” is not accurate for ChatGPT as a whole. Prism3 does not offer paid advertising, so I cover ads no further than saying they are a different system.

What OpenAI does not publish

OpenAI’s pages give no weights for any factor and do not name a particular data field that moves you up. Its merchant page does say that richer, current feed data helps products surface with more accurate information, which is a general statement, not a ranking rule. Its web search article says ChatGPT ranks results using multiple factors and that placement is not guaranteed, and its product feed documentation says eligibility does not guarantee display.

That is the edge of the documented world; everything beyond it is inference. (This post stays on product selection; for how ChatGPT picks sources to cite in general, see how to rank in ChatGPT.)

Claims you will see that I cannot back up

I looked for each of the following in the OpenAI and Shopify pages above and did not find them:

• That ChatGPT’s product carousel is simply Google Shopping data, with a precise percentage attached. Vendor studies make this claim, but OpenAI’s pages name structured provider metadata and, for shopping research, publicly available retail sites. I have not verified any such study at its primary source, so I am not repeating a figure.
• That click-through, dwell time or other shopper behaviour on product cards moves products up or down.
• That review sentiment is weighted in a particular way in the ranking.
• That ChatGPT systematically favours premium products over practical ones. I have not read a primary study behind this and I am not repeating it.

Some may be true, and OpenAI can change how it works at any time. Decision rule: act on an input only if an official page names it, or if fixing it is good for your customers anyway. Complete descriptions and honest reviews pass both tests. Chasing an unverified behavioural signal passes neither. Our myths guide covers the common ones.

Where Shopify fits

If your store is on Shopify, the product data side is handled for you, with conditions. OpenAI’s help article says product data for Shopify merchants is already integrated into ChatGPT through Shopify Catalog with no additional work required, and its merchant page says the same for Shopify and Etsy stores. Shopify’s help pages add:

• ChatGPT is a discovery-focused referrer. Shoppers finish in your own online store checkout, in an in-app browser or a new tab, so there is no ChatGPT direct checkout setting in your admin. OpenAI’s merchant page agrees: it says it is moving away from a standalone Instant Checkout experience, with purchases completed on merchant websites or apps.
• The channel is active by default for eligible stores and managed under Sales channels > Agentic. There are no fees beyond standard payment processing fees.
• You can remove ChatGPT’s access through Shopify Catalog. Shopify says it can take up to 7 days to stop, and that your products might still be referenced through crawling and indexing.

Shopify’s ranking signals are Shopify’s, not ChatGPT’s

Shopify’s help page for managing agentic storefronts has a search preview tool. It runs queries against Shopify Catalog and shows a Listing quality indicator for each product, built from five signals: description completeness (word count), image coverage, product reviews (average rating and count, calculated only from reviews verified by trusted sources), variant and option completeness (including whether option names use hard-to-read acronyms or numbers) and shop policy completeness.

This is how Shopify scores listings in its own Catalog search. Shopify says agentic storefronts often re-rank results with their own logic, that the preview may differ from what customers see, and that it is a directional signal, not an exact prediction. It also says relevance depends on other signals such as popularity, customer engagement and brand recognition.

Shopify’s own guidance says optimising your products can increase the chances of inclusion in AI platforms’ results, and also that other factors specific to each platform influence them. The five signals make a sensible checklist because they describe complete, trustworthy listings. They are not OpenAI’s ranking factors, and nobody outside OpenAI can honestly tell you how far a higher Shopify score moves your ChatGPT position.

What you control: four gates

Think of the controllable parts as four gates, in order. A product that fails an early gate never reaches the later ones.

Gate 1: Are your products eligible at all?

Check these first, because no copywriting fixes an ineligible store.

• Where you sell. OpenAI’s merchant page says shopping is currently live for ChatGPT users in the US, with more regions planned. Shopify’s ChatGPT page says your store must sell to customers in the United States, although the store can be based outside it. Australian founders should read that twice: if you ship only within Australia, you are outside the condition Shopify states for this channel. Our guide to AI shopping in Australia covers what each assistant says about Australia, and why I would not open a US market only to qualify.
• Shopify Catalog requirements. Shopify says your store must be on the Starter plan or higher and not in private mode, and your products must have a title and at least one image, a price above zero, publication to your online store and an identifiable product URL. Products cannot have Unlisted status or be hidden from search engines. Shopify also reviews account standing over time, including a track record of genuine sales and a low chargeback rate.
• Your policies and terms. Shopify says you need your Terms of service, Privacy policy and Return and refund policy completed under Settings > Policies to sell on ChatGPT, and that you must read and agree to the Shopify Agentic Storefronts Supplemental Terms of Service.
• Not B2B-only. Shopify says B2B-only products are not supported, and products behind a login or password are excluded when Shopify can identify them. If an app or theme edit hides prices until login, Shopify says it may not detect that and the products might still show; it suggests options such as hiding specific products from AI channels.
• Product categories. Shopify lists categories not allowed in Shopify Catalog, such as alcohol, tobacco and vaping, CBD supplements and services. OpenAI’s commerce policies, updated 29 September 2026, have their own prohibited list, which includes supplements: ingestible products marketed for weight loss, sexual enhancement, bodybuilding, detoxification or other health-related effects. Shopify adds that products on OpenAI’s prohibited list do not display in ChatGPT product discovery results. If you sell wellness products, read both lists against your range. This is not legal advice; if you are unsure whether a product is covered, ask OpenAI or Shopify.

Where to look: Sales channels > Agentic. Shopify says it shows how many of your products are in Shopify Catalog, or an Ineligible message that explains why.

Gate 2: Can ChatGPT read your product data correctly?

This is the gate with the most room to work. Shopify says the product details AI platforms consider include title, description, images, product type, vendor, collections, tags, barcodes (ISBN, UPC, GTIN and so on) and variants including option names. It also notes that other factors specific to each platform influence results.

OpenAI’s developer documentation shows what its feed format holds. It has nine required fields: item ID, title, description, product URL, brand, seller name, main image, availability and price. Optional fields include GTIN, MPN, material, colour, size, dimensions, weight, shipping price, return acceptance and policy link, and product review count and average rating. Direct feed onboarding is, as at 30 October 2026, available to approved partners, and OpenAI tells Shopify merchants no additional work is required, so do not build a feed. Use the list as a picture of what the system can hold.

These pages do not say which Shopify fields map to which OpenAI fields, so I will not pretend to. A store with a plain title, a factual description, clean variants, a barcode where one exists, an accurate price and honest stock status simply gives any system, ChatGPT included, less to get wrong. Start with your ten best-selling products: fix the products that earn the money before the products that are easy to fix.

Three further data controls sit in Shopify’s Channel mapping page for Shopify Catalog:

• Product fields. If your real title or description lives in a metafield, you can point Shopify Catalog at it. If the chosen source is empty for a product, Shopify falls back to the default.
• Variant grouping. If separate products are really variants of one product, for example tag-prefixed or delimiter-separated titles, you can group them so they show as one listing.
• List price. You can choose whether Shopify Catalog sends the price or the compare-at price as the list price. Shopify says AI agents might use the list price to show strikethrough pricing. Whether a compare-at price is honest is a consumer-law question, not a data question. The ACCC said in a 16 September 2016 media release that “was/now” price representations are likely to be misleading if the product has not been sold at the “was” price for a reasonable period immediately before the sale. Check the ACCC’s current guidance or a professional before you rely on strikethrough pricing. This is not legal advice.

Shopify’s ChatGPT page also says to include any relevant legal disclosures in the first 6,000 characters of the description. For regulated goods, get the wording from someone qualified. The product page SEO checklist goes field by field.

Gate 3: Do your policies and reviews give a reason to trust you?

Policies. Shopify says to keep store policies complete and up to date so that, when a customer asks an AI agent about your returns, the agent can reference what you stored in Shopify. Its knowledge base, under Sales channels > Agentic, holds non-product facts such as your return process and shipping timelines, and Shopify says agents can query the published version through its Storefront MCP. Shopify never changes it without your approval. Read each fact against reality, and make sure a public page on your store says the same thing. Shopify’s pages do not say how ChatGPT uses the knowledge base when choosing products, so treat it as accuracy insurance, not a ranking lever. Our guide to shipping and returns messaging for Australian stores covers the wording. OpenAI’s commerce policies also tell merchants to avoid hidden fees and misleading refund or return policies.

Reviews. OpenAI says its review summaries come from reviews on public websites and are not verified by OpenAI. Shopify’s listing quality counts rating and review count from reviews verified by trusted sources. Two different systems, one direction: genuine reviews that are easy to find on your product page and on public sites. OpenAI’s commerce policies list fake reviews and fabricated endorsements as prohibited, so do not buy shortcuts. Our guide to product reviews and star ratings on Shopify covers how to collect and show them.

Gate 4: Can OpenAI’s crawler read your site?

OpenAI’s crawler page says OAI-SearchBot is used to surface websites in ChatGPT’s search features. Sites that opt out of it will not be shown in ChatGPT search answers, though they can still appear as navigational links. OpenAI recommends allowing OAI-SearchBot in robots.txt and allowing requests from its published IP ranges, and says a robots.txt change can take about 24 hours to take effect. GPTBot, the crawler for model training, is a separate setting. The shopping pages I read do not name the crawler behind shopping research, so allowing OAI-SearchBot is the documented step, not a guarantee.

For Shopify stores, Shopify says network-layer bot management is handled for you. What you can touch is /robots.txt, through the robots.txt.liquid template. Shopify adds that robots.txt rules are advisory, and that blocking AI crawlers affects only open-web discoverability, not the product data Shopify Catalog sends to channels you have switched on.

On 30 October 2026 I fetched the default robots.txt of a Shopify demo store. It named no OpenAI crawlers and did not block product pages. That says nothing about your store, which may carry edits from a theme, an app or a previous agency. Open yourstore.com/robots.txt and look for OAI-SearchBot, GPTBot or a blanket Disallow: /. If a rule is there that you did not write, find out who did before deleting it. Our robots.txt guide explains what is safe to change.

What you do not control

Knowing this stops a lot of wasted effort. You do not control the ranking weights (not published), the shopper’s context (budget, memory, custom instructions, clarifying answers), other merchants’ prices (which decide “Best price” labels), the model-written titles, labels and review summaries, or whether you are shown at all, because not every available product is shown and eligibility is not a guarantee.

The one-hour check, in order

1. Sales channels > Agentic. Note whether Shopify Catalog shows a product count or an Ineligible reason, and the date.
2. Settings > Policies. Confirm Terms of service, Privacy policy and Refund policy are complete.
3. Markets. Check whether your store sells to US customers. If not, read our Australia guide (linked above) before changing anything.
4. Hidden or gated products. Look for Unlisted products, password-protected sections and apps that hide prices behind a login.
5. Top ten products. Run the six-row table above, product by product.
6. Channel mapping. Open Review product data mapping and check that the title and description sources are what your product page shows, and decide the list price setting.
7. robots.txt. Look for blocks on OAI-SearchBot.
8. Knowledge base. Read each published fact against reality.
9. Search preview. Run ten queries a buyer would type and note which products appear and the Listing quality notes. Treat it as directional, as Shopify says.
10. Reviews. Check that your genuine reviews appear in readable text on the product page.

Change things in small batches and write the date next to each, or you will not know what worked. Our agentic storefronts guide covers the Shopify side in more detail.

How to test without fooling yourself

Opening ChatGPT and asking for your product is a biased test: OpenAI says the query and the user’s context, such as memory and custom instructions, shape results. A fairer routine:

• Write ten prompts a real buyer would use, such as “best [product type] for [need] under [budget]”, not your brand name.
• Run them in a clean browser profile or a second account with no relevant memory, and again in your usual one.
• Log whether you appeared, whether price and availability were right, whether shipping and returns details were right, and where the link went.
• Repeat monthly and note any change you made in between.

The accuracy checks matter more than “appeared?”: each wrong detail points to a source you can fix.

Measuring orders from ChatGPT

Shopify says orders from AI channels show in your admin with channel or referrer attribution, and Sales channels > Agentic reports sales, orders, online store sessions and online store conversion for each AI channel you sell on. For GA4, see tracking ChatGPT traffic in GA4 on Shopify.

Keep expectations proportionate. As an illustration, not a benchmark: 12 ChatGPT sessions in a month of 2,000 is 0.6% of sessions. A handful more or fewer next month swings that share widely, so judge it over several months.

Key takeaways

• OpenAI does not publish how ChatGPT ranks products. Treat any formula, percentage or “secret signal” as unverified unless an OpenAI or Shopify page states it.
• OpenAI documents relevance to the shopper’s intent and context, structured provider metadata, safety and product policies and, for merchant lists, availability, price, quality and maker or primary seller.
• OpenAI says product results are not ads. ChatGPT also has separate, labelled ads that OpenAI says do not influence answers.
• For Shopify stores, product data reaches ChatGPT through Shopify Catalog. Shopify’s listing quality score is Shopify’s signal, not OpenAI’s.
• As at 30 October 2026, OpenAI says shopping is live for US users and Shopify says the ChatGPT channel needs a store that sells to US customers. Australian founders should check their own admin first.
• You control eligibility, product data, policies, reviews and crawler access. Work in that order, best sellers first.
• Test with a fixed prompt log from a clean account and measure orders over months, not days.

FAQs

Can I pay to rank higher in ChatGPT shopping?

OpenAI says product results are selected independently, are not ads and are not influenced by OpenAI partnerships. ChatGPT does have ads, but OpenAI says they are separate, labelled and do not influence answers. I found no documented way to buy a better product-result position.

Do I need to submit a product feed to OpenAI?

Not if you are on Shopify. OpenAI’s help article and merchant page say Shopify merchants’ product data is already integrated, with no application needed. Its developer documentation says direct feed onboarding is available to approved partners as at 30 October 2026.

Is ChatGPT shopping available in Australia?

Not confirmed. OpenAI’s merchant page says shopping is currently live for ChatGPT users in the US, and its shopping help article lists no countries. Shopify’s ChatGPT page says your store must sell to customers in the United States. Check Sales channels > Agentic in your own admin, and read our Australia guide.

Does ChatGPT shopping just copy Google Shopping?

No OpenAI or Shopify page I read says so. Some vendor studies claim a Google Shopping link, but I have not verified any at the primary source, so I do not repeat their figures.

Why does ChatGPT show a competitor and not my product?

OpenAI does not say, and several reasons are possible: your listing may match the stated need less clearly, your data may be thinner, your store may be ineligible in that country, another merchant may be the maker or primary seller, or the shopper’s memory and context may have favoured something else. Run the one-hour check, then the prompt log, and change one thing at a time.

Are reviews important for ChatGPT shopping?

OpenAI says review summaries are generated from reviews on public websites, that reviews and ratings are not verified by OpenAI, and that reviews are among the things ChatGPT may consider. It does not say how much weight they carry. Genuine, visible reviews are worth having for your shoppers regardless.

Want a second pair of eyes on your Shopify store?

Our GEO service, sold alongside SEO, covers structured data and schema markup, citation-worthy content formatting and monitoring AI search visibility. We do not guarantee citations, placement or sales from any AI assistant. If you would like a second opinion on your store, book a free 30-minute diagnostic or tell us about your store: we look at your store, and at what the AI assistants say about your category, before we recommend anything. No pitch.