Product Schema for AI: What to Implement and What Is Theatre

If you own your storefront, add Schema.org Product markup and keep a clean feed: identity fields (GTIN, brand, MPN, model) first, then price, availability and shipping. If your product lives on a marketplace, you cannot add schema to the page. The page is built from the fields you fill in, so your structured attributes are your schema, and every empty field is a property that ships blank. Adding markup to a page you do not control is theatre. Filling attribute fields, and keeping them consistent with your own site and feed, is not. Below, one real Amazon listing is mapped field by field to Schema.org Product, so you can see what you can and cannot control.

What agents actually read

Nobody outside the platforms can tell you which signals each AI shopping surface reads. Amazon has not published how Alexa for Shopping (formerly Rufus) weights its sources, and most assistants are the same.

Google has documented its side. Its Search Central page on product structured data states: "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."

That is the honest basis for this article. Machine-readable product facts, consistent across places, are documented to help. Anything beyond that is a guess.

Feed, schema and listing fields: three different things

A marketplace seller owns the third row only. That tells you where the work is.

If you own the site: what to implement

Start with the properties that identify the product, then the commercial ones.

→ Identity: name, brand, gtin (or gtin12, gtin13), mpn, model, sku.

→ The thing itself: color, material, size, weight, image, description.

→ Commercial: offers, with price, currency and availability.

→ No schema name for it: additionalProperty, which Schema.org defines as a property-value pair for "a product feature or another characteristic for which there is no matching property in schema.org".

Then send the same facts in a feed and make sure both agree with the page.

If you sell on a marketplace: your attributes are your schema

You cannot paste JSON-LD into an Amazon detail page. You control the fields Amazon gives you, and each one lines up with a Schema.org property or has none.

The HTML I fetched for the listing below, on 7 October 2026, contained no JSON-LD block. There is nothing on that page for you to add to. The field table is the only structured data you author, so treat every field as a property you are filling for a machine reader.

A real ASIN, mapped field by field

The product is the STANLEY Quencher ProTour Flipstraw Tumbler, 30 oz, Emerald Honeydew, ASIN B0H1YLJZFF on amazon.com. I read its page on 7 October 2026 and the Schema.org Product definitions at schema.org/Product the same day. Only fields I saw are listed.

Three things stand out.

→ Thirteen of the sixteen fields are yours. Nine map to a named property. Capacity, special feature, uses and included components fall into the catch-all additionalProperty.

→ Three fields are not yours. Reviews belong to customers. The ASIN and the sales rank belong to Amazon. No amount of effort moves them.

→ Some fields are missing. The page text I fetched showed no item weight row. Schema.org has weight, height, width and depth. On this listing those properties would ship blank.

A site owner could mark up the same product like this. It uses only the fields above, with no price, because I did not capture one.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "STANLEY Quencher ProTour Flipstraw Tumbler, 30 oz, Emerald Honeydew",
  "brand": { "@type": "Brand", "name": "STANLEY" },
  "gtin12": "041604516948",
  "mpn": "100000146856",
  "model": "100000146883",
  "material": "Stainless Steel",
  "color": "Emerald Honeydew",
  "additionalProperty": [
    { "@type": "PropertyValue", "name": "Capacity", "value": "30 fluid ounces" },
    { "@type": "PropertyValue", "name": "Special feature", "value": "Vacuum Insulated" }
  ]
}

What is theatre

→ Adding markup to a page you do not control. You cannot edit a marketplace detail page, so there is nothing to mark up.

→ Markup volume. A long list of vague properties is worse than a short list of true ones.

→ Markup that disagrees with the page. Schema describes the page. It does not replace it.

→ Schema as a ranking promise. The Google page I read talks about eligibility and verification, not ranking.

Not theatre: the identity fields, a valid GTIN, and a feed that matches your page.

Consistency: the failure nobody checks

Missing data leaves a question unanswered. Contradictory data is worse, because a system that cannot tell which source is right has no safe answer to give.

This listing shows how it starts: two UPCs against one ASIN. That can be legitimate, for example if the ASIN groups several variants. Know which is which, and make sure your own site and feed use the matching one.

Check across every place the product appears:

→ Title, brand and size are worded identically.

→ The GTIN is the same everywhere and matches the variant.

→ Capacity, material and color agree in every field. "30 oz" in the title and "32" in an attribute is a contradiction.

The wider pattern is in the multi-surface visibility guide.

What to do next

1. Count your empty attribute fields. Open the attribute list for your marketplace category and mark every blank.
2. Fill identity first. GTIN, brand, model, material, dimensions.
3. Compare the listing with your own site. Write down every contradiction and fix the source that is wrong.
4. If you own a storefront, add Product markup there and keep it identical to the marketplace data.
5. Stop adding markup to pages you do not control.

Where to start

A free Cosmy account at cosmy.ai runs an Alexa for Shopping analysis on one of your ASINs and returns a gap report of the buyer questions your listing does not answer. It does not write schema. It shows where your listing leaves facts out.