Shopify schema for AI search: make product data agree

The decision in brief

• Do not add a second Product entity until you know what the theme and apps already render.
• Map price, availability, currency, variant, brand, identifier, review, and shipping facts to an accountable source.
• Use Shopify's structured data output as a baseline only when it matches the store's actual data and visible page.
• Validate both feature eligibility and vocabulary. Google's test and Schema.org Validator answer different questions.
• Measure search visibility and AI referrals separately. Correct schema can improve interpretation, but it does not prove causation.

Four layers that must agree

The practical failure is usually not a missing schema type. It is a disagreement between systems. This matrix shows where to inspect each fact and what to do when two layers conflict.

Choose one output owner before changing code

1. Inventory every emitter. Inspect rendered HTML for Product, ProductGroup, Offer, Review, AggregateRating, Organization, and BreadcrumbList. Record whether each came from the theme layout, product template, app embed, app block, custom code, or another integration.
2. Compare facts, not entity counts. Match the selected variant, URL, SKU or GTIN, price, currency, availability, image, brand, ratings, shipping, and returns against the visible page and Shopify catalog.
3. Keep the accurate maintained owner. If one theme or app already emits complete, current data, correct its mapping. Do not add a second Product entity merely because another tool reports missing optional fields.
4. Replace only with a rollback path. When the current owner is inaccurate or unmaintained, save its output and configuration, disable one emitter, publish the replacement, validate the rendered page, and keep a tested restoration step.

Decision rule: use the implementation that produces one coherent rendered product graph from accountable source fields. “App” and “custom” are maintenance labels, not proof of quality.

Two safe implementation paths

Keep the existing owner when the theme or a maintained app already emits one accurate Product entity. Fix the underlying Shopify fields or that owner's mapping instead of adding another graph.

Use a theme-owned baseline when the theme does not emit Product data and the team can maintain Liquid. Shopify documents the structured_data filter for supported objects. A product-template pattern is:

{% raw %}{% if product %}
  <script type="application/ld+json">
    {{ product | structured_data }}
  </script>
{% endif %}{% endraw %}

This is a baseline, not permission to duplicate output. Before release, inspect whether the theme layout, product template, app embed, or app block already creates the entity. For stores with product variants, compare the rendered model with Google's current merchant-listing guidance and Schema.org's ProductGroup vocabulary before extending it.

Worked example: one product group, two purchasable variants

This fictional example shows the relationship to inspect; it is not a copy-and-paste promise of Google eligibility. The catalog is the source for the product group, variant URLs, SKUs, GTINs, prices, currency, and availability. The visible page must substantiate the same selected-variant facts.

{
  "@context": "https://schema.org",
  "@type": "ProductGroup",
  "@id": "https://store.invalid/products/harbor-tee#group",
  "name": "Harbor Tee",
  "productGroupID": "HARBOR-TEE",
  "variesBy": ["https://schema.org/size"],
  "hasVariant": [
    {
      "@type": "Product",
      "name": "Harbor Tee - Small",
      "url": "https://store.invalid/products/harbor-tee?variant=small",
      "sku": "HT-S",
      "size": "S",
      "offers": {
        "@type": "Offer",
        "price": "32.00",
        "priceCurrency": "USD",
        "availability": "https://schema.org/InStock",
        "url": "https://store.invalid/products/harbor-tee?variant=small"
      }
    },
    {
      "@type": "Product",
      "name": "Harbor Tee - Large",
      "url": "https://store.invalid/products/harbor-tee?variant=large",
      "sku": "HT-L",
      "size": "L",
      "offers": {
        "@type": "Offer",
        "price": "32.00",
        "priceCurrency": "USD",
        "availability": "https://schema.org/OutOfStock",
        "url": "https://store.invalid/products/harbor-tee?variant=large"
      }
    }
  ]
}

The .invalid host marks this as a non-live teaching example. Replace it only with the real canonical product and variant URLs. GTIN is deliberately omitted: if a valid identifier does not exist, do not invent one. Test whether the store's current theme and Google requirements support the chosen variant model before release.

Mismatch and rollback ledger

Five mechanisms that should not be collapsed into “AI schema”

Validation and rollback checklist

• Capture before: save the rendered JSON-LD and validation results for a simple product, a multi-variant product, a sale item, and an out-of-stock item.
• Name the owner: record whether theme, app, feed, or catalog owns each changed fact.
• Change one emitter: avoid simultaneous theme and app changes that make the source of a regression unclear.
• Compare facts: verify name, URL, image, currency, price, availability, SKU or GTIN, brand, variant, shipping, returns, and review provenance against the live page.
• Run both validators: use Google Rich Results Test for documented Google features and Schema.org Validator for vocabulary and graph inspection.
• Retain rollback: keep the previous snippet or app configuration and the release date so a conflicting emitter can be restored quickly.

Use priceValidUntil only when an offer actually expires on that date. Do not create ratings, GTINs, availability, or shipping facts to make a validator look complete.

Primary documentation and what each source decides

• Shopify Liquid: structured_data filter documents the platform's generated JSON-LD output.
• Google Product structured data documents Product snippet requirements and recommendations.
• Google merchant listing structured data covers commerce properties such as offers, shipping, returns, and variants.
• Schema.org ProductGroup, Product, and Offer define the underlying vocabulary.
• Google AI features and your website says standard search technical requirements apply and no special AI file or markup is required.
• Shopify: optimize your store for AI and agentic storefront product settings describe Shopify's current controls.
• OpenAI merchant guidance documents the separate merchant-product discovery path.