ChatGPT Shopping: How Your Products Appear in the Chat
Where do ChatGPT shopping results get products from? A practical ecommerce guide to product feeds, OAI-SearchBot, product page data and traffic tracking.
ChatGPT shopping is when a user searches for products in a ChatGPT conversation, compares options with images and prices, and goes to the seller’s site to buy. There are two ways for your products to appear in these results: sending a product feed to OpenAI, or making sure your product pages can be read by ChatGPT’s search crawler. For an ecommerce site in Turkey, the real work today is the second path, because the feed-based experience is currently limited to users in the US.[1]
In this post I explain where ChatGPT finds products, how a feed differs from web crawling, which information must be complete on a product page, and how I measure the traffic that comes in. Because I come from a software development background and manage advertising for a Shopify-based store, I put extra weight on the technical side.
How does ChatGPT shopping work?
ChatGPT shopping works when a user describes a need in a natural sentence and ChatGPT shows matching product options. According to OpenAI’s merchant page, shoppers browse products in one place through results enriched with images, prices and key details.[1] Products are ranked for each person based on their preferences, the context of the conversation and the product data.[1] The purchase itself is completed on the seller’s own website or app.
OpenAI’s developer documentation describes this journey in four steps: the customer describes the need, ChatGPT looks for relevant options, the customer compares offers, and the journey moves to the browser or an integration.[2]
The most important consequence of this flow for an ecommerce business is this: ChatGPT is not the party that makes the sale, it is the party that shows the product. The user discovers the product in the chat, but comes to your product page to add it to the cart. So every bit of visibility gained in ChatGPT ultimately depends on the quality of the product page on your own site.
The user can also narrow things down over several rounds. When they ask for “the lighter one”, “the one with faster shipping” or “the one with a longer return window”, ChatGPT updates the list against the new criteria. Even if your product appears in the first list, it can be dropped at the second question. If dimensions, warranty, shipping time or return terms are not written on your page, the model cannot evaluate you on those questions.
Where does ChatGPT find products?
ChatGPT can get product information from three kinds of sources: product feeds sent by merchants, other retail sources and the public web. OpenAI’s feed guide clearly states that ChatGPT can use feeds and these other sources for product discovery.[2] Which source carries more weight depends on the user’s question and on the integrations the merchant has set up.
Putting the two main paths side by side makes it clear which job to start first:
With a product feed
- You send product data to OpenAI in a structured format
- Price, stock and image freshness stay under your control
- Identity verification and country selection are required
- Currently US-focused; Turkey cannot be targeted
- Shopify and Etsy catalogs are already integrated
With web crawling
- ChatGPT reads your product page with OAI-SearchBot
- The information is whatever the page says
- No extra application is needed
- The path open to stores in Turkey today
- If the crawler is blocked, the page is not shown as a source
There is one point to watch on the feed side. In OpenAI’s feed schema, the target_countries and store_country fields are required. According to the documentation, the shared format currently represents the US, Canada and Mexico, and standard uploads are treated as US regardless of this field.[3] For a store that only sells domestically, this means a feed does not reach Turkish shoppers today.
Behind the feed infrastructure is a standard called the Agentic Commerce Protocol (ACP). According to OpenAI’s merchant page, the Agentic Commerce Protocol is an open foundation for AI-powered commerce in ChatGPT, and today it supports product discovery through shared product data.[1] OpenAI says the protocol will cover the whole shopping journey over time through deeper integrations.[1] This direction is called agentic commerce. Agentic commerce is a model in which an AI agent researches products on the user’s behalf and carries out permitted steps; the user always gives the approval.
How does a store in Turkey appear in ChatGPT?
A store in Turkey today appears in ChatGPT mostly through product and category pages cited as sources in search answers. The prerequisite is allowing access to OAI-SearchBot, OpenAI’s search crawler. According to OpenAI’s crawler documentation, sites that block OAI-SearchBot are not shown in ChatGPT search answers and can only appear as navigational links.[4]
OpenAI uses separate bots for training and search. GPTBot crawls content for training generative models. OAI-SearchBot works to surface sites in ChatGPT’s search features.[4] The two are managed with separate lines in robots.txt. A setup that is closed to training but open to search looks like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /
I covered this distinction and other AI bots in a table in my AI crawlers and robots.txt guide. What matters here is that the search bot stays open for shopping visibility.
Category pages also matter on this path. For a broad question such as “a blender for a small kitchen”, ChatGPT may cite pages that list and compare products. A category page with filters, criteria and a short guide text answers these questions better. I explained how to build one in detail in my category page SEO post.
Which information must be complete on a product page?
A product page should carry, as text, every piece of information an AI tool needs to answer the user’s question. That means product name, brand, model, dimensions, material, compatibility, price, stock, shipping time and return terms. A size chart written inside an image, or technical details that only load when a tab is opened, may not be readable by a crawler.
OpenAI’s feed schema is a good checklist of which information is considered important. Nine fields are required in the schema: item_id, title, description, url, brand, seller_name, image_url, availability and price.[3] Titles are recommended to be at most 150 characters and descriptions at most 5,000 characters; the description should be plain, factual text.[3] Even if you do not send a feed, your product page should give a clear answer to each of these fields.
| Information | Role in ChatGPT | Check question |
|---|---|---|
| Product name and brand | Correct product matching | Is the same name used on the site, marketplaces and Merchant Center? |
| Price | Filter in budget questions | Is the visible price the same as the price in structured data? |
| Stock and variant | Whether the option can be bought | Does a sold-out size or color still show as “in stock”? |
| Dimensions, material, compatibility | Answers to narrowing questions | Is this information in the text rather than in an image? |
| Shipping and returns | Decision information in comparisons | Are return window and shipping time written on the product page? |
| Product reviews | Review summary and trust | Are reviews published as text on the page? |
The feed schema also sets a clear rule for reviews. The review_count and star_rating fields must describe the same review population, and store reviews must not be mixed with product reviews.[3] The same rule applies to the product page: summing reviews shared across variants again for each variant inflates the rating and damages trust.
Product and Offer structured data on the page (schema.org markup) repeats this information in a machine-readable form. Google Search Central recommends adding Product and Offer markup to product pages.[5] OpenAI does not explain in detail how ChatGPT uses this markup. Even so, keeping price, stock and brand identical on the page, in the markup and in the feed gives the product a consistent identity across every channel. I covered the required feed attributes on the Google side in a table in my Merchant Center free listings post.
How do you submit a product feed to ChatGPT?
Short answer
How do you submit a product feed to ChatGPT?
To submit a product feed to ChatGPT, you claim your business in OpenAI’s Business Hub and verify your identity. You then select the countries and currencies you sell in, and send a schema-compliant feed by manual upload, a hosted HTTPS URL or SFTP. As the final step, you fix the upload errors.
- Claim your business through Business Hub.
- Complete identity verification in the products section.
- Select the countries and currencies you sell in.
- Prepare a file that follows the feed schema with all required fields.
- Choose manual upload, a hosted HTTPS URL or SFTP.
- Fix the errors in the upload report.
OpenAI’s feed guide describes the setup in these steps and states that the feed is delivered as a full snapshot.[2] The feed does not update a cart or confirm a purchase. When the customer acts, the current offer is still checked on the live site or integration.[2]
OpenAI’s merchant page also gives two practical facts. Stores selling through Shopify or Etsy already have their catalogs integrated and need no extra setup or application.[1] Other merchants join a waitlist through the application form; OpenAI says it is building a merchant platform for self-serve data integration.[1]
- Clean up the catalog
Use a stable ID for every product and variant. List variants in separate rows and link them with a shared group ID.
- Fill in the required fields
Fill in the nine required fields and the country fields in every row. Do not write placeholders such as null or n/a for unknown values.
- State stock explicitly
Use one of in_stock, out_of_stock, pre_order, backorder or unknown in the availability field. A row left empty is rejected.
- Set the update frequency
Plan how quickly price and stock changes reach the feed, based on your campaign calendar.
The schema sets a clear rule for the stock field: an empty, missing or unrecognized availability value causes the row to be rejected.[3] Not putting the price in the product ID is another explicit warning in the schema; the ID should stay the same when price or stock changes.[3]
Why do price and stock consistency matter?
Price and stock consistency is the precondition for ChatGPT to show your product with correct information, and for the user not to be disappointed on arriving at your site. OpenAI’s merchant page recommends syncing price and stock updates with the merchant’s own systems.[1] If a product shown as discounted in the chat turns out to be more expensive on the site, the user will most likely leave before reaching the cart.
The most common cause of inconsistency is campaign periods. The discount changes instantly on the site, but the feed updates once a day or the markup on the page still carries the old price. The same problem leads to product disapprovals in Google Merchant Center; on the ChatGPT side, it creates the risk of appearing with wrong information.
On the software side, my recommendation is to feed the catalog not as a scheduled file but with an update triggered whenever price or stock changes. That way the site, the feed and the markup change at the same time. Testing this flow before big campaigns is also part of Black Friday SEO preparation.
How do you measure traffic from ChatGPT?
Traffic from ChatGPT is measured by tracking visits with chatgpt.com as the session source in GA4 as a separate segment. You report which product and category pages this segment lands on, along with add-to-cart and purchase events. First confirm in the traffic source report how the source name appears in your own data.
This measurement has two limits. First, you only see users who click through to your site. A user who sees your product in ChatGPT and later searches for your brand on Google is recorded as a brand search. Second, some visits from the mobile app may not carry referrer information and can look like direct traffic.
That is why I track three signals together:
- Referral traffic: chatgpt.com sessions in GA4, their landing pages and purchase events.
- Brand search: how queries containing the brand name change over time in Search Console.
- Question set test: regularly asking ChatGPT 10-20 real shopping questions from your category and recording whether your products and source pages appear.
For measurement to be reliable, consent and server-side infrastructure also need to be set up correctly. I explained these steps in my server-side tracking post.
Where should you start with ChatGPT shopping today?
The starting point for ChatGPT shopping visibility is not opening a new channel, but making your existing product pages readable by AI tools. This work is a natural extension of classic ecommerce SEO. I explained the general framework for appearing in AI answers in my what is GEO post; the list below is that framework adapted to ecommerce.
If you use Shopify or Etsy, the catalog integration already exists; your job is to improve the quality of the product data. If you use another platform and sell to the US, a merchant application and feed preparation are a sensible next step. If you only sell within Turkey, focus on crawler access and product pages before setting aside a separate budget for a feed.
Conclusion: ChatGPT shows the product, the product page makes the sale
ChatGPT shopping moves part of product research into the chat window, but the sale is still completed on the seller’s site. For stores in Turkey, the most effective path today is product pages that are open to OAI-SearchBot, complete in their information and consistent in price across every channel. When the feed-based experience becomes more widely available, the same clean data will be ready for that channel too.
As a next step, look at OAI-SearchBot requests in your server logs and check your ten best-selling product pages against the list above. I describe how product data, measurement and advertising fit together on my SEO services and performance marketing pages.
Key terms
- ChatGPT shopping
- ChatGPT shopping is a product discovery experience in which a user searches for products in a ChatGPT conversation, compares options with images, prices and key details, and is sent to the seller's site to buy. Product data can come from merchant feeds and the public web.
- Agentic Commerce Protocol
- Agentic Commerce Protocol (ACP) is the protocol OpenAI describes as an open foundation for AI-powered commerce in ChatGPT. Today it supports product discovery through shared product data; OpenAI says it plans to extend it over time with deeper integrations and APIs to cover the whole shopping journey.
- Agentic commerce
- Agentic commerce is a commerce model in which an AI agent researches products on the user's behalf, compares options and, where allowed, carries out purchase or booking steps. The user makes the decision and gives the approval.
Frequently asked questions
Can ChatGPT shopping be used in Turkey?
Users in Turkey can ask ChatGPT about products and see links to web sources in the answers. However, according to OpenAI's merchant page, the shopping experience based on product feeds is currently available only to users in the US. For a store in Turkey, the main path today is making product pages readable by OAI-SearchBot.
Do I have to pay for my products to appear in ChatGPT?
No. According to OpenAI's merchant page, ChatGPT sends shoppers to your own website or app by default. There is no fee for purchases completed on your site. Ads are a separate product with their own feed requirements.
Does my Shopify store need to apply separately for ChatGPT?
OpenAI's merchant page says that stores selling through Shopify or Etsy already have their catalogs integrated and need no extra setup. Still, the accuracy of product titles, descriptions, stock and prices remains your responsibility.
Where does ChatGPT find products?
According to OpenAI's developer documentation, ChatGPT can use merchant product feeds, other retail sources and the public web for product discovery. Which source is used depends on the request and the integrations that are available.
How do I measure sales from ChatGPT?
In GA4, you can track traffic with chatgpt.com as the session source as a separate segment. Report which product pages this segment lands on and its purchase events. First check how the source name appears in your own data; this measurement only shows users who click through.
Sources
The full topic: E-commerce SEO guide ↗ · GEO guide ↗