Which Schema.org Markup AI Shopping Agents Rely On Most

The schema.org markup AI shopping agents rely on most, going by what Google, Microsoft and OpenAI put in writing, comes down to a short stack: Product with a nested Offer carrying price, currency and availability, alongside a description, a brand and a stable ID; then the global identifiers (gtin, mpn); then shipping and return policy; then ratings. That order reflects how many of the three platforms name each item and how firmly they word it. We found no platform that publishes a weighting, so treat it as a tally of documentation and a sensible sequence for the work.

What the three platforms publish, side by side

Schema.org markup for AI shopping agents is documented unevenly. Google publishes required and recommended properties for product pages. Microsoft publishes a recommended field list on the Bing Webmaster Blog, plus separate microdata rules in its advertising help for stores using Microsoft Merchant Center. OpenAI publishes a product feed specification, and in its commerce documentation we found no schema.org requirement for product pages. The table uses each vendor's own status words.

Name, image and price form the floor: two sources require them and the third recommends them. Availability, description, brand, URL and a stable ID come next, since OpenAI requires them while Google and the Bing post recommend or welcome them. Global identifiers and shipping follow, then return policy and ratings.

What Google requires for merchant listings

Google's product structured data for Product, Offer, review, shipping and return policy markup in merchant listings follows schema.org, and the reference page was last updated on September 8, 2026. Google Search Central lists three required Product properties: name, image and offers. Inside the Offer, a price and a priceCurrency are required, and "merchant listing experiences require a price greater than zero." Everything else is recommended, including availability, itemCondition, shippingDetails and hasMerchantReturnPolicy on the offer, and description (which Google calls strongly recommended), brand, sku, mpn, the gtin family, aggregateRating and review on the product.

Shipping and returns bring their own required sub-properties once you add them. For the shipping details enhancement, OfferShippingDetails needs deliveryTime, shippingDestination and shippingRate. MerchantReturnPolicy needs applicableCountry and returnPolicyCategory, plus merchantReturnDays when the return window is finite. Google recommends declaring a business-wide policy under Organization markup and keeping offer-level properties for products that differ from it.

Two technical guidelines sit above any single property. Only pages where a shopper can buy the product qualify for merchant listing experiences, and Google recommends Product markup in the initial HTML, warning that dynamically generated markup "can make Shopping crawls less frequent and less reliable" for fast-changing content such as availability and price. Pages that review a product without selling it fall under product snippets, which the Google Search Central overview describes as having more options for review information.

For its AI surfaces Google asks for no extra markup. The Google Search Central guide to AI features, last updated December 10, 2025, says of AI Overviews and AI Mode: "There's also no special schema.org structured data that you need to add." Among the SEO basics it still calls worthwhile are structured data that matches the visible text on the page and Merchant Center information that is up to date.

How Microsoft documents Product markup for Bing and Copilot

Bing's product structured data guidance for schema.org Product markup, the kind Bing Shopping and AI agents read on product pages, sits across four Microsoft documents. The general one comes from Bing Webmaster Tools, whose guidelines cover "Bing search experiences, Copilot, and grounding API results" and address structured data briefly: "Structured data may support clearer grounding but does not guarantee visibility or grounding traffic. Markup must accurately reflect visible content." The same guidelines describe generative engine optimization as a focus on content eligibility for grounding and reference in AI responses.

The property list comes from a Bing Webmaster Blog post dated May 19, 2025, about IndexNow for shopping. It recommends title (name in JSON-LD), description, price, link, image, shipping ("especially important for Germany and Austria"), a unique id, brand, gtin, mpn, datePublished and dateModified, with category, seller and itemCondition as optional additions. Its sample markup is JSON-LD, which the Bing Webmaster Tools markup help page lists as a supported schema.org format next to HTML Microdata, adding that "Our crawlers do not prefer one specification over another." On feeds, the blog post says that by implementing IndexNow with that schema "you can skip the need for a separate merchant feed" (a claim tied to adopting IndexNow).

Microsoft's merchant documentation reads differently. A Microsoft Advertising help page on Microsoft Learn, last updated July 27, 2026, states: "Although it's possible to markup HTML using RDF and JSON, Bing only supports Offer and Product microdata." The microdata "must be embedded in the HTML returned from your web server" and must also be static and match the values shown to the user. We found no current Bing Webmaster page that sorts Product properties into required and recommended the way Google's does. If your store runs through Microsoft Merchant Center, confirm which format your offers are read from before you retire microdata.

ChatGPT shopping is specified as a feed

Essential schema markup for an AI shopping assistant looks different at OpenAI, whose commerce documentation specifies a product feed and, in the pages we read, sets no schema.org requirement for product pages. The OpenAI product reference asks for nine required fields per item: item_id, title, description, url, brand, seller_name, image_url, availability and price. GTIN and MPN are optional in the OpenAI format, as are the shipping charge, the return fields and the review aggregates. The same page warns that "Eligibility does not guarantee display."

Status matters before anyone plans around this. The OpenAI get-started guide says "Onboarding product feeds in ChatGPT is currently available to approved partners" and the product reference says checkout "requires a separately enabled integration." For a store outside that program the field list still works as a checklist, since eight of the nine required fields have a counterpart among the Product and Offer properties in Google's reference; seller_name is the one without. Product data is also only part of the picture: the wider work of getting cited by ChatGPT applies to stores too.

One product page, annotated

The product page schema.org properties for AI agents covered above, as structured data current to the 2026 documentation, fit in one JSON-LD block. The example below is hypothetical and every value in it is a placeholder. If you are assembling a block like this by hand, our schema generator is a further resource.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Tarn 20L Daypack",
  "description": "20-liter daypack in recycled nylon with a padded laptop sleeve.",
  "image": ["https://example.com/img/tarn-20-grey-1x1.jpg"],
  "sku": "TARN-20-GRY",
  "gtin13": "9506000134352",
  "mpn": "TRN20G",
  "brand": { "@type": "Brand", "name": "Example Outfitters" },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": 4.6,
    "reviewCount": 212
  },
  "offers": {
    "@type": "Offer",
    "url": "https://example.com/products/tarn-20-grey",
    "price": 89.00,
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition",
    "shippingDetails": {
      "@type": "OfferShippingDetails",
      "shippingRate": { "@type": "MonetaryAmount", "value": 0, "currency": "USD" },
      "shippingDestination": { "@type": "DefinedRegion", "addressCountry": "US" },
      "deliveryTime": {
        "@type": "ShippingDeliveryTime",
        "handlingTime": { "@type": "QuantitativeValue", "minValue": 0, "maxValue": 1, "unitCode": "DAY" },
        "transitTime": { "@type": "QuantitativeValue", "minValue": 2, "maxValue": 5, "unitCode": "DAY" }
      }
    },
    "hasMerchantReturnPolicy": {
      "@type": "MerchantReturnPolicy",
      "applicableCountry": "US",
      "returnPolicyCategory": "https://schema.org/MerchantReturnFiniteReturnWindow",
      "merchantReturnDays": 30,
      "returnMethod": "https://schema.org/ReturnByMail",
      "returnFees": "https://schema.org/FreeReturn"
    }
  }
}

Reading it from the top:

• name, image and offers with price and priceCurrency are the whole of Google's required set.
• description, brand and availability are recommended by Google and required in OpenAI's feed. Keep availability tied to live stock, because Google and Microsoft both ask in writing that markup match what the shopper sees.
• sku, gtin13 and mpn are the identifiers. Google says to "Include all applicable global identifiers" and wants GTINs in numerical form.
• shippingDetails carries the three sub-properties Google requires for its shipping enhancement, and a value of 0 declares free shipping.
• hasMerchantReturnPolicy sits on the offer to keep the example in one block. Google's preference is an Organization-level policy, with offer-level markup for exceptions.
• aggregateRating is recommended by Google, and OpenAI's optional review_count and star_rating carry the same aggregate into a feed.

We found no published weighting

We found no platform that publishes a weighting of schema.org properties for AI shopping answers in any of the ten pages cited here. The closest statements are general ones. Google's overview advises that "it is recommended to provide as much rich product information as available, without concern for the exact experiences that will use it." The OpenAI key concepts page says recommended attributes such as rich media, reviews and performance signals "improve ranking, relevance, and user trust." It attaches no numbers.

The order at the top of this page is therefore a tally of documentation, and schema by itself secures neither a rich result nor a recommendation on any of the three platforms. Where the documents converge is on agreement between sources. Google states that "Providing both structured data on web pages and a Merchant Center feed maximizes your eligibility to experiences and helps Google correctly understand and verify your data." When shipping or return policies are defined in several places, product-level Merchant Center feeds take precedence and organization-level markup ranks last. A feed that states one price while the page states another is the failure worth designing against, for the same reason brand facts have to agree everywhere an assistant might look.

Whether assistants then name your products is a separate measurement. Track it as AI share of voice across the shopping prompts your buyers use, and treat markup completeness as the input you control. Our AI visibility checker is a further resource for a first reading, and AI search optimization is the service we run around that measurement.

Sources

• Google Search Central: Merchant listing (Product, Offer) structured data
• Google Search Central: Introduction to Product structured data
• Google Search Central: AI features and your website
• Bing Webmaster Tools: Bing Webmaster Guidelines
• Bing Webmaster Blog: IndexNow Enables Faster and More Reliable Updates for Shopping and Ads
• Bing Webmaster Tools: Marking Up Your Site with Structured Data
• Microsoft Learn: Using structured data markup
• OpenAI: Products (feed reference)
• OpenAI: Get Started
• OpenAI: Key concepts