If you sell products online, you have probably seen the news that ChatGPT now shows products with photos, prices and buy links. The big question is how ChatGPT shopping picks which products to show, and whether yours can be one of them. The short answer: ChatGPT chooses products itself, using product data and reviews from third-party sources. You cannot pay for a spot, but you can make your products easier to find and describe.
OpenAI announced the update on 28 April 2025, so it is only two days old as I write this. In this guide I explain what has changed, how products get shown, what OpenAI has and hasn’t said, and the checks I’d make on an online store this week.
What Is ChatGPT Shopping?
ChatGPT shopping is a set of product features inside ChatGPT search. When you ask something with shopping intent, such as “best running shoes for flat feet under $120”, ChatGPT can now answer with a small set of product cards instead of plain text.
Each card shows a product image, a short title, a price and a rating. If you click a product, a detail view opens with more information, a summary of what reviewers say, and a list of stores that sell it. Each store has a link that takes you to its own site to buy.
Here is what OpenAI shared at launch:
- Who gets it: Plus, Pro and Free users, and people who use ChatGPT without logging in. OpenAI says it is rolling out in all regions where ChatGPT is available, and that it may take a few days to reach everyone.
- When it appears: only when ChatGPT thinks your question is about shopping. A normal question still gets a normal answer.
- Starting categories: reports from the launch say OpenAI is starting with fashion, beauty, home goods and electronics.
- No ads: OpenAI says the product results are chosen independently and are not ads.
- No checkout in ChatGPT: you don’t buy inside ChatGPT. The buy links send you to the merchant’s website, where the normal checkout happens.
The same update added a few other search features. ChatGPT search now works in WhatsApp, answers can show more than one citation for a single point, and the search box shows trending searches and autocomplete suggestions. OpenAI also said people made over a billion web searches in ChatGPT in the past week. That number tells me this is no longer a side feature.
How ChatGPT Shopping Chooses Which Products to Show
This is the part that matters most for store owners. OpenAI’s help documentation explains the selection in general terms. It hasn’t published a ranking guide, and I don’t expect one soon. But the inputs are clear enough to act on.
Where the Product Data Comes From
ChatGPT builds its product results from structured metadata (organised product details such as price, description and reviews) supplied by third-party providers. OpenAI hasn’t listed who all those providers are.
OpenAI also says ChatGPT writes its own simplified product titles and descriptions from that data. So the text on the card may not match the title on your product page word for word.
What ChatGPT Weighs When It Picks Products
According to OpenAI, ChatGPT looks at things like:
- Price, and price comes first if the user gives a budget.
- Customer ratings and reviews from around the web.
- The user’s own needs, such as size, colour, material or use case.
- Earlier messages in the chat, so a follow-up question builds on what came before.
- Memories and custom instructions the user has saved in ChatGPT.
- OpenAI’s safety standards, which can keep some products out.
Notice what is missing: there is no bid, no sponsored slot and no “promote this product” button. OpenAI has also said it does not earn a commission on purchases made through these results. That means the only way in is to have good, clear product data that matches what people ask for.
Labels, Ratings and Review Summaries
Some cards carry short labels like “Budget-friendly” or “Most popular”. OpenAI says ChatGPT creates these labels itself, and they are not guarantees.
Star ratings and review counts are pulled together from third-party providers. That is why the rating on a card can differ from the rating on your own site. The written review summary is based on reviews published on public websites.
In my experience, this is where many stores are weakest. If your product has a handful of reviews that only live inside a widget on your site, ChatGPT has very little to summarise.
How the Store List and Prices Are Set
When a product is sold by several stores, the detail view lists them. OpenAI says the order of that list is mostly set by the data providers. It is not sorted by lowest price or best shipping.
OpenAI also warns that prices can lag behind changes on a merchant’s site, and that taxes and delivery costs may differ at checkout. The first price shown isn’t always the lowest one. If your price changes often, there is a real chance ChatGPT shows an old number for a while.
| What shows in ChatGPT | Where it comes from, per OpenAI | What you can influence |
|---|---|---|
| Product title and description | Written by ChatGPT from third-party product data | Clear, specific product names and descriptions on your pages and feeds |
| Price | Third-party providers, sometimes with a delay | Accurate prices in your page markup and any product feeds you run |
| Star rating and count | Combined from third-party providers | Collecting genuine reviews on platforms that are publicly visible |
| Review summary | Reviews on public websites | Reviews that crawlers can actually read |
| Labels like “Most popular” | Generated by ChatGPT | Not directly; they follow the data |
| Store list order | Mostly set by the data providers | Not directly at this stage |
Can Merchants Submit Products to ChatGPT Shopping?
Not yet. OpenAI says it is exploring a way for merchants to send product feeds directly, for more accurate and current listings. For now, there is an interest form merchants can fill in to be told when submissions open. OpenAI hasn’t given a date.
I’d fill in the form if you run an online store. It costs nothing, and being early to a new product channel rarely hurts. But don’t wait for it. The work that helps you today is the same work that will help when feeds open.
If you already run a Google Merchant Center feed, keep it clean and current. OpenAI hasn’t said it uses Google’s data, so I’m not suggesting a link. The point is simpler: the habits that keep a Shopping feed healthy, like accurate prices, stock status and identifiers, are the same habits any product data system rewards.
How to Prepare Your Product Pages for ChatGPT Shopping
Here is the order I’d work in. None of this is new or risky. It is good ecommerce SEO that now has one more reason behind it.
1. Let OpenAI’s Search Crawler Reach Your Store
ChatGPT search uses a crawler called OAI-SearchBot to find and show websites. OpenAI’s separate crawler, GPTBot, collects content for training its models. You can control them separately in robots.txt (the file that tells crawlers which parts of your site they may visit).
Some stores added a broad block on OpenAI’s crawlers a while ago and never looked at it again. If that is you, you may be blocking search as well as training. A setup that allows search but blocks training looks like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /Whether you block GPTBot is a business choice. I just want it to be a choice you make on purpose. My robots.txt guide covers the syntax and the mistakes I see most often.
Also check your firewall, CDN and bot protection settings. When I review a store, this is the first thing I look at, because a security tool can block a crawler even when robots.txt allows it.
2. Put Product Facts in Plain HTML
Crawlers read the HTML your server sends. If your price, stock status or specs only appear after JavaScript runs, some systems may miss them.
Open “view source” on a product page and search for the price. If you can’t find it there, talk to your developer. This is a common problem on headless stores and some page builders.
3. Add Complete Product Structured Data
Structured data is code that labels the facts on a page, such as “this is the price” or “this is the rating”. Most online stores already add Product markup for Google, using the schema.org vocabulary.
OpenAI hasn’t said it reads this markup directly. But it says its results rely on structured product data, and clean markup is the most reliable way to publish your product facts in a form machines understand. I’d make sure each product page includes:
- The product name, brand and a clear description.
- An Offer with price, currency and availability.
- Identifiers such as GTIN (the barcode number), MPN or SKU, where you have them.
- Review and rating data, but only for reviews that genuinely appear on the page.
- Shipping and return details if you sell directly.
Google’s product structured data documentation is the best reference I know for the fields and rules. If you want to understand when star ratings are allowed, I wrote a separate guide on review snippets and star ratings.
4. Write Product Descriptions That Answer Real Questions
People ask ChatGPT long, specific questions. “Waterproof hiking boots for wide feet that work in snow” is a typical example. Products with vague descriptions are hard to match to questions like that.
For each important product, I check that the page clearly states:
- Who the product is for and what it is best at.
- Exact sizes, dimensions, materials and compatibility.
- What makes it different from the cheaper or pricier option next to it.
- Care, warranty and delivery details.
Write these in plain sentences, not only in a spec table. A clear line like “fits up to a 16-inch laptop” is easy for any system to match to a question.
5. Build Reviews That Live on the Open Web
ChatGPT summarises reviews from public websites. So reviews need to be visible and crawlable, not hidden behind a login or loaded only by a script.
I’d ask happy customers for reviews in a steady, honest way, and reply to negative ones calmly. Reviews on well-known retailer sites, marketplaces and review platforms also add to the picture. Never buy or fake reviews. Apart from the ethics, it is exactly the kind of signal these systems are built to discount.
6. Keep Prices and Stock Accurate
OpenAI openly says prices shown can be out of date. You can’t fix their delay, but you can avoid making it worse. Update your markup and feeds when prices change, and mark out-of-stock items correctly.
In my experience, stale “in stock” labels on sold-out products cause more lost trust than almost any other data error. A shopper who clicks through and can’t buy rarely comes back.
How to Track Traffic From ChatGPT Shopping
When someone clicks a store link in ChatGPT, they land on your site as referral traffic. In GA4, open the Traffic acquisition report and look for chatgpt.com as the session source.
OpenAI hasn’t shared a dashboard for merchants, so you won’t see impressions or rankings. What you can do:
- Create a GA4 comparison or exploration filtered to sessions where the source contains “chatgpt”.
- Look at which landing pages get those visits. Product pages there are a sign you are being shown.
- Track revenue and conversion rate for that traffic separately from Google organic.
- Note the 28 April launch date so you can compare before and after.
Expect small numbers at first. My rule of thumb with any new channel is to watch it for a few months before drawing conclusions.
What We Don’t Know Yet
It is early, and there are real gaps in what OpenAI has shared. I think it helps to be honest about them:
- Which data providers are used. OpenAI says “third-party providers” but hasn’t named them all.
- How ranking works in detail. We know the inputs, not the weights.
- When merchant feeds open. There is only an interest form so far.
- Whether ads will come later. The current results are not ads, and OpenAI says it earns no commission. Its leaders have talked publicly about possible future models, but nothing has been announced for shopping.
- How memory changes results. OpenAI says ChatGPT will start using memory to personalise shopping for Plus and Pro users. It also says this won’t be available in the EU, UK, Switzerland, Norway, Iceland or Liechtenstein.
Personalisation is worth thinking about. If results depend on each user’s chat history and memories, two people asking the same question may see different products. That makes “ranking first” a less useful goal than being a clear, well-described option.
What ChatGPT Shopping Means for Ecommerce SEO
I don’t think this replaces Google Shopping or product search on Google any time soon. But it adds a new place where buying decisions start. It also follows a clear trend. OpenAI has opened ChatGPT search to everyone, even without a login, and is testing agents like Operator that browse and act on websites.
For most stores, the right reaction is not a new strategy. It is to finish the basics properly: open access for search crawlers, product facts in HTML, complete structured data, honest reviews and accurate prices. Stores that already do this well are in the best position.
The one mindset change I’d suggest is this. With no ad slot to buy, your product data is your pitch. If ChatGPT can’t tell what your product is, who it is for and what it costs, it has no reason to recommend it.
Conclusion
ChatGPT shopping shows products chosen by ChatGPT, not paid ads, using third-party product data, prices and public reviews. You can’t buy a place in it, and merchant feeds aren’t open yet. What you can do is make your products easy to crawl, easy to understand and easy to trust.
If you only do one thing this week, check your robots.txt and firewall rules for OAI-SearchBot. Then work through your top product pages, starting with structured data and the price shown in your HTML.
