Price and stock as labels: product data in plain words

Published 30 October 2026 · 4 min read

Structured data is a set of hidden labels that tell machines your price, stock and rating. What they do for AI, what they don't, and how to check yours.

A buyer asks ChatGPT for a waterproof hiking boot under €150. Next to the answer, ChatGPT can show a row of products with pictures and prices.

Nobody typed those prices in by looking at your page. They come from data: facts about each product, written down so a machine can read them without guessing. If your pages don't carry that data, or carry it wrong, your product has less to go on when an assistant picks what to show.

The data has an off-putting name, structured data. Here it is in plain words: what it is, what it does for you, what it doesn't, and how to check your own shop in ten minutes.

Labels hidden in the page

Every product page has two audiences. People see the photo, the price in large type and a green “in stock”. Robots read the page's code, where “€129” is just a number among other numbers.

Structured data is a set of labels hidden in that code. They say: this is the product's name, this is its price and currency, this is whether it's in stock, this is its rating from this many reviews. The labels follow a shared vocabulary called schema.org. For a shop, two of its terms matter: Product, the thing you sell, and Offer, the price and stock you sell it at. Visitors never see any of it.

This is roughly what the labels look like. You won't need to write them yourself; it just shows how plain they are:

{
  "@type": "Product",
  "name": "Ridge Wide hiking boot",
  "offers": {
    "@type": "Offer",
    "price": "129.00",
    "priceCurrency": "EUR",
    "availability": "https://schema.org/InStock"
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.7",
    "reviewCount": "38"
  }
}

How the labels travel to ChatGPT

OpenAI says the products ChatGPT shows when people shop come from structured metadata, supplied by merchants and by third-party providers. It doesn't say who those providers are.

Google looks like one of them. A study by Search Engine Land and Peec (March 2026) found 83% of the products in ChatGPT's shopping carousel among strong matches in Google Shopping's top organic results. Google Shopping reads these labels from shop pages. It also takes a product feed, a list of all your products with prices and stock, from Google Merchant Center, Google's tool for shops.

So the labels on your pages, and a Merchant Center feed if you have one, are how your price and stock get beyond your own site. OpenAI also says it ranks the shops in that carousel partly on price and availability, one more reason for those two facts to be right.

What labels can't do

Labels help. They don't decide on their own who gets recommended.

Much of what decides lies off the page: whether search engines have the shop indexed, links from other sites, and being on the rankings and comparison sites that ChatGPT's searches turn up. Get the labels right because it's quick and it helps. Then spend your effort there too.

Check your labels in ten minutes

Most shop platforms add labels automatically, so there's a good chance yours are already there. Here's how to see what they say.

1. 1 Open Google's Rich Results Test at search.google.com/test/rich-results and paste the address of one of your best-selling product pages.
2. 2 Look at what it found. You want “Product snippets” or “Merchant listings” on the list. If neither is there, the page has no product labels.
3. 3 Open the result and look for four things: the product's name, a price with its currency, availability, and a rating if the product has reviews.
4. 4 Compare them with the page as a customer sees it. Same price? Same stock? A label saying €129 on a page showing €119 leaves any reader unsure which to believe.
5. 5 Repeat for a product that's sold out and one that's on sale. Those are where labels most often fall out of date.
6. 6 If labels are missing or wrong, ask your platform's support or your developer. It's often a setting or the shop's theme.

The words on the page count as much as the labels. In our own test, an in-stock product's labels correctly said “in stock”, but the shop's template kept a hidden “sold out” box on every page. In all three answers, ChatGPT said the page showed contradictory stock and the buyer should contact the shop first. If your template keeps a hidden “notify me when available” box, ask your developer to leave it out of pages where the product is in stock.

One more question for whoever looks after the shop's code: are the labels in the page as the server sends it, or added later by a script? Google's test runs scripts. A study by Vercel found that the robots fetching pages for AI don't.