Structured product data for AI shopping agents

Short answer: put a schema.org Product in JSON-LD on every product page, with an Offer that states the price, currency and stock status, plus an image, a brand and at least one identifier (gtin, sku or mpn). If the product comes in sizes or colours, publish it as a ProductGroup with one Product per variant. AI shopping agents read this data to compare, recommend and buy; without it they have to guess from your page layout.

Check your product pages with the free agent-readiness audit. Structured data is the biggest single category in the score.

Why structured data matters more for agents than for search

A search engine can still rank a page with patchy markup. An agent comparing ten stores for “waterproof trail boots, size 9, under £120” needs to know, for certain, the price, currency, whether size 9 is in stock and what the product is. If any of that is missing or only shown visually, the agent either skips your store or gives the shopper a wrong answer.

The fields that matter

Field	Where	Why agents need it
name	Product	Matching the shopper’s request
description	Product	Understanding what the product is
image	Product	Showing the product back to the shopper
brand	Product	Brand searches and comparisons
gtin / gtin13, sku, mpn	Product	Identifying the exact item across stores
price	Offer	Comparing and filtering on price
priceCurrency	Offer	A price without a currency is ambiguous
availability	Offer	Not recommending things that are out of stock

Price, currency, availability and an identifier are the most important. A GTIN (the barcode number) is best because it identifies the same item at every retailer.

A minimal complete example

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Ridge waterproof trail boot",
  "description": "Leather trail boot with a waterproof membrane and a Vibram sole.",
  "image": ["https://yourstore.example/images/ridge-boot.jpg"],
  "brand": { "@type": "Brand", "name": "Ridgeline" },
  "sku": "RIDGE-BRN",
  "gtin13": "5012345678900",
  "offers": {
    "@type": "Offer",
    "url": "https://yourstore.example/products/ridge-boot",
    "price": "119.00",
    "priceCurrency": "GBP",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition"
  }
}

Put it in a <script type="application/ld+json"> tag in the page’s HTML. It must be in the HTML the server sends, not added later by JavaScript, because many agents don’t run scripts.

Sizes and colours: use ProductGroup

A dropdown of sizes is invisible to most agents. Describe each variant as its own Product inside a ProductGroup:

{
  "@context": "https://schema.org",
  "@type": "ProductGroup",
  "name": "Ridge waterproof trail boot",
  "productGroupID": "RIDGE",
  "variesBy": ["https://schema.org/size", "https://schema.org/color"],
  "brand": { "@type": "Brand", "name": "Ridgeline" },
  "hasVariant": [
    {
      "@type": "Product",
      "name": "Ridge waterproof trail boot, brown, UK 9",
      "sku": "RIDGE-BRN-9",
      "size": "UK 9",
      "color": "Brown",
      "offers": { "@type": "Offer", "price": "119.00", "priceCurrency": "GBP", "availability": "https://schema.org/InStock" }
    },
    {
      "@type": "Product",
      "name": "Ridge waterproof trail boot, brown, UK 10",
      "sku": "RIDGE-BRN-10",
      "size": "UK 10",
      "color": "Brown",
      "offers": { "@type": "Offer", "price": "119.00", "priceCurrency": "GBP", "availability": "https://schema.org/OutOfStock" }
    }
  ]
}

Each variant has its own sku, price and availability, so an agent can answer “is it in stock in a 10?” correctly.

Platform notes

WooCommerce

WooCommerce outputs basic Product JSON-LD, but variable products usually list a price range rather than each variation, and GTINs are often missing. Add the GTIN field (WooCommerce has a built-in “GTIN, UPC, EAN or ISBN” field in recent versions) and use an SEO or schema plugin that outputs ProductGroup and hasVariant for variable products.

Magento / Adobe Commerce

Many Magento themes use microdata rather than JSON-LD, and configurable products keep their options in swatch JavaScript. Add JSON-LD with an extension or in the product view template, and output each child product as a variant.

Custom stores

Generate the JSON-LD on the server from the same data that renders the page, so price and stock can never disagree with what the shopper sees.

Common mistakes

• Price without currency, or a price formatted as text such as "£119". Use "price": "119.00" and "priceCurrency": "GBP".
• Availability as plain text such as "In stock". Use the schema.org URL, for example https://schema.org/InStock.
• Invalid JSON, often from an unescaped quote in a product description, which makes the whole block unreadable.
• Markup added by JavaScript after the page loads.
• Stale data: a cached page that still says InStock after the item has sold out.

How to test it

Run the free agent-readiness audit. It samples up to five product pages and reports which fields are missing on how many of them, whether variants are machine-readable, and whether any JSON-LD block fails to parse. For a single page, Google’s Rich Results Test and the Schema.org validator are also useful.