Fix the catalog
Product schema for AI shopping: what matters
By Ankit Minocha, founding team at Atomz. Updated January 21, 2026.
Structured data is how an assistant reads a product without guessing. Here is which schema fields move the needle, which platforms lean on them, and how to add them on Shopify.
Product schema is the structured-data layer that states, in a format a machine reads directly, what a product is: its name, price, availability, brand, and the attributes that define it. A shopper reads your page and infers all of that from layout and copy. An assistant cannot, so it reads the schema, and when the schema is thin or missing it either guesses or skips the product entirely.
How much this matters depends on the platform, which is the part most guides skip. Gemini leans hardest on schema because it draws from Google's Shopping Graph, where the feed and the structured data are the product as far as the system is concerned. ChatGPT works through the Agentic Commerce Protocol and a feed, so schema matters but the feed carries more weight. Perplexity sits in the middle, reading the structured content it crawls. Claude leans on the surrounding context. The common thread is that none of them reads your design, so the more of the truth you encode as data, the more answerable you are.
The fields that earn the recommendation
The core fields are the obvious ones, name, price, availability, brand, and a clear product type, and you should never ship a product page without them. The fields that win a recommendation are the ones that answer intent, and on schema.org those live in additionalProperty.
{
"@type": "Product",
"name": "Daily Vitamin C Serum",
"brand": { "@type": "Brand", "name": "Example Co" },
"additionalProperty": [
{ "@type": "PropertyValue", "name": "Skin type", "value": "Oily" },
{ "@type": "PropertyValue", "name": "Concern", "value": "Dark spots" },
{ "@type": "PropertyValue", "name": "Active", "value": "Vitamin C 15%" },
{ "@type": "PropertyValue", "name": "Fragrance", "value": "Unscented" }
]
}
That block is the difference between a product an assistant can recommend for 'a fragrance-free vitamin C serum for oily skin' and one it cannot, because the query maps to those properties exactly.
The point
Core fields make a product valid, but the additionalProperty fields are what make it findable. A page can pass every schema validator and still be invisible to a real shopper question, because validation checks format, not whether you encoded the attributes a shopper searches on.
Adding it on Shopify
Since the Spring 2026 Edition, Shopify also reads your structured data into Shopify Catalog and will infer the attributes you leave blank, with varying accuracy, so the schema you write is what replaces the guess.
On Shopify you have two practical routes. You can edit the product template in your theme to output the JSON-LD from your product and metafield data, which keeps the schema in sync with the catalog. Or you let a tool write the structured attributes to your metafields and emit the schema for you, which is the path that scales when you have thousands of SKUs and no appetite for editing Liquid by hand. Either way, the schema is only as rich as the attributes behind it, so the real work is filling those attributes first. The audit shows which of your products carry enough structure to be recommended and which read as blanks.
Subscribe to The Agentic Operator