Digital Product Passport and AI: the product data shopping agents need

ChatGPT, Microsoft Copilot and AI Mode in Google Search show products right in their answers. What they say about a product comes from merchant data: product feeds, catalogs such as Shopify Catalog, and the product page. The feeds ask for GTIN, brand and material, and shoppers also ask about origin and care. These are the details that belong in a Digital Product Passport. If you prepare your data for the passport, you also have it ready for AI shopping agents.

Key points

  • In agentic commerce, AI agents research, compare and select products for shoppers.
  • The agents read product data from feeds, from catalogs and from the structured data on the product page.
  • OpenAI and Google ask for GTIN, brand and material as separate fields. Google strongly recommends the GTIN.
  • A Digital Product Passport keeps these details structured and machine-readable at a permanent address.
  • Shopify, Shopware, WooCommerce and JTL all have fields for this data.

What is agentic commerce?

Agentic commerce means AI agents research, compare and select products on behalf of shoppers. Shopify describes it as a shopping model in which AI agents make decisions across the entire shopping journey (Shopify, September 25, 2026).

AI shopping agents now sit inside ChatGPT, Microsoft Copilot, AI Mode in Google Search and the Gemini app. Shopify passes its merchants’ products on to these services (Shopify, June 17, 2026). Buying inside the chat runs on dedicated protocols. Shopify and Google developed the open Universal Commerce Protocol, and OpenAI documents the Agentic Commerce Protocol for ChatGPT. Both define how an agent makes a purchase in a store. What the agent knows about the product still depends on the product data.

What product data do AI agents need?

Agents get product data in three ways.

Product feeds. OpenAI documents a product feed that merchants use so ChatGPT can find and display their products. Nine fields are required, including title, description, brand, price and availability. Optional fields cover GTIN, material, color, size, dimensions and weight. The GTIN needs a valid check digit. If a product has none, OpenAI says not to invent a manufacturer part number as a substitute. The feed has no field for origin, care or certifications.

Google asks for similar data in Merchant Center. The GTIN is optional there, but Google strongly recommends it. Material is required when variants differ in it. There is also a field for product details with up to 100 entries and, in the EU, a field for certifications.

Catalogs. Shopify collects its merchants’ products in Shopify Catalog. The catalog assigns categories, extracts attributes and consolidates variants. Merchants group their products for this through standard attributes and metafields (Shopify, December 10, 2025). According to Shopify, eligible products are included in the catalog by default. When a product shows up in AI conversations but doesn’t sell, Shopify’s assistant Sidekick suggests filling in missing product details (Shopify, June 17, 2026). Neither post says which countries this covers.

The product page. Google reads structured data based on schema.org on the product page. For merchant listings, Google recommends GTIN, brand, material, color, size and up to ten certifications, among others. According to Google, AI Overviews and AI Mode need no special markup. Structured data should match the visible text on the page, and Merchant Center information should be up to date.

What does the Digital Product Passport have to do with it?

An agent can only answer a question as precisely as the data allows. Here are questions shoppers ask about a product, and the passport field that answers each one:

QuestionField in the passport
Is this sweater pure merino wool?Material with fiber percentages
Where was this jacket sewn?Origin with country of manufacture
Can this shirt go in the dryer?Care instructions
Is this exact model available in size M?GTIN of the variant
Who made this product?Manufacturer with address
Does the fabric carry a label?Certificates

The passport is based on open standards and is, where appropriate, machine-readable and structured (Art. 10(1)(d) ESPR). Which details become mandatory is chosen by the delegated act for each product group from a list in Annex III of the ESPR. That list includes the GTIN or an equivalent identifier, instructions for use and information about the manufacturer. More on what the regulation requires is in What is a Digital Product Passport? (in German). For textiles, the fiber composition must already be shown in your store before purchase (Art. 16(1) Regulation (EU) 1007/2011).

An agent states what is in the data. If someone looks for a pure merino sweater and your product has no material field, the agent only has the description text. If the material isn’t there either, the agent can’t name it. If a competitor has filled in the field, the agent can recommend that product with the matching detail. None of the providers named here has published whether agents prefer products with a passport. What is documented is that OpenAI, Google and Shopify ask for these details as fields, and the same details belong in the passport.

The same applies between businesses. Once a delegated act applies to a product group, retailers make the manufacturer’s passport easily accessible to their customers, including in their online store (Art. 31(2) ESPR). If the passport sits at a permanent address in structured form, a retailer can take the details into their store and feeds without retyping them.

For details from the store, the feed and the passport to end up with the right product, each product needs a unique identifier. A GTIN identifies one item variant, for example a T-shirt in size M and white. OpenAI and Google both provide for it in their feeds, and the ESPR names it as a possible identifier in the passport. When you need your own GTIN and what it costs is covered in GTIN for the Digital Product Passport (in German).

GS1 Digital Link turns the GTIN into a web address that a QR code on the label opens. Under the standard, the address can point to more than one source: the passport page for people and structured data for machines. A shopper with a phone and a program collecting product data both reach the same product. The format isn’t mandatory, but it is widely used. Details are in GS1 Digital Link explained (in German).

What does machine-readable product data look like?

Our sample passport sits at an address based on GS1 Digital Link. When a program requests JSON-LD there, it gets the passport as structured data. The brand and manufacturer are made up. An excerpt:

{
  "@context": "https://schema.org",
  "@type": "Product",
  "@id": "https://mapasso.de/01/09524524558003",
  "productID": "09524524558003",
  "inLanguage": "en",
  "name": "Kiesel Zip Hoodie",
  "brand": { "@type": "Brand", "name": "Mapasso Demo" },
  "manufacturer": {
    "@type": "Organization",
    "name": "Mapasso Demo",
    "address": "Musterstraße 1, 10115 Berlin, Deutschland"
  },
  "countryOfOrigin": "PT",
  "material": "Outer fabric: 100 % cotton; Organic cotton: 100 % of the outer fabric; Ribbing: 95 % cotton, 5 % elastane; Organic cotton: 95 % of the ribbing; Hood lining: 100 % cotton; Organic cotton: 100 % of the hood lining",
  "additionalProperty": [
    {
      "@type": "PropertyValue",
      "name": "careInstructions",
      "value": "Wash at 40 °C, inside out, with similar colors. Do not bleach. Do not tumble dry. Iron at medium temperature. Undyed cotton may shrink slightly the first time you wash it."
    }
  ]
}

The field names come from the schema.org vocabulary. Brand and material carry the same names as in Google’s recommendations for product pages. The GTIN is part of the address: in GS1 Digital Link, the 01 after the slash stands for GTIN. Origin and care have their own fields. A program doesn’t have to pull them out of running text.

On the product page in your store, the same details look different. There they appear as text for people. Depending on the theme, some of them also appear as structured data in the page source.

What can merchants on Shopify, Shopware, WooCommerce and JTL do now?

All four systems have fields for GTIN and material:

SystemGTINMaterial and other details
ShopifyBarcode with type GTIN on the variantCustom product metafield, shown on the product page through a dynamic source in the theme
Shopware 6GTIN/EAN field on the productProperties appear on the product page, custom fields stay internal
WooCommerce 9.2 and laterGTIN, UPC, EAN or ISBN on the product and on each variationAttributes appear in the store, custom fields only once the theme outputs them
JTL-WawiGTIN in the item master dataCharacteristics, sent to JTL-Shop at the next sync

In Shopware, material, origin and care therefore belong in properties if shoppers and search engines should see them. Custom fields suit proof that stays in the background. How to set up the details in each system is covered in the guide Preparing textile data for the Digital Product Passport.

What you can do now:

  1. Enter the GTIN for every variant if you have one. Don’t enter numbers that aren’t yours.
  2. Set up material, origin and care as separate fields and show them on the product page.
  3. Check whether your theme or feed plugin writes these fields into the structured data and the product feed.
  4. Keep Merchant Center and the product page in sync. Feed, product page and passport should draw on the same data, or you end up with three different details in three places.
  5. Create a passport with a permanent address for your products. Then all details sit in one place, for shoppers through the QR code and for machines in structured form.

Digital Product Passports with Mapasso

Mapasso is software for Digital Product Passports. It creates the passport from the data already in your store or ERP, with a QR code and an address based on GS1 Digital Link. The same address serves the passport as JSON-LD for machines, as in the excerpt above. For Shopify, the app is available in the Shopify App Store. For any other online store or ERP, there is the Mapasso portal at portal.mapasso.de, where you upload your products as a CSV or Excel file. The portal interface is in German. You enter passport content in all 24 EU languages.

Frequently asked questions

Do I need a product passport for AI to find my products?

No. AI agents read feeds, catalogs and product pages. Whether you need a passport depends on your product group and its delegated act. The data overlaps, though: agents ask for GTIN, brand, material and certifications, and these also belong in the passport.

How do products get into ChatGPT Shopping?

OpenAI documents a product feed that merchants use so ChatGPT can find and display their products. Shopify merchants with eligible products are in Shopify Catalog by default and turn on ChatGPT in the Agentic Storefronts area (Shopify, June 17, 2026). OpenAI and Shopify don’t say in these sources which countries this covers.

Does Google read structured product data?

Yes. With product markup based on schema.org, a page can appear in merchant listings on Google Search, for example in the shopping knowledge panel. Google doesn’t guarantee it. The GTIN is one of the recommended properties, and Google advises including all applicable identifiers.

Sources (16)

All accessed on November 11, 2026.