Product Schema Markup Specification for AI Shopping Citations

TL;DR

Use schema.org Product for every commerce-intent product page. Choose between the merchant-listing variant (when the user can buy on your page) and the product-snippet variant (when they cannot, e.g. editorial reviews). Include name, image, description, brand, sku, gtin or mpn, an Offer with price/availability, and AggregateRating plus Review when first-party reviews exist. For multi-variant catalogs, use ProductGroup with hasVariant. Keep schema in sync with visible page text and Google Merchant Center feed; the 2026 Merchant Center update enforces image and video attributes that should be reflected on-page.

Two Variants: Merchant Listing vs Product Snippet

Google distinguishes two Product markup classes depending on whether users can purchase on the page (Google Search Central — Merchant listing structured data).

• Merchant listing. The user can buy from this page. Required fields are stricter: Offer.price, Offer.priceCurrency, Offer.availability, plus shipping and return policy. Eligible for shopping knowledge panel, Google Images merchant rich results, popular product results, and ChatGPT Shopping crawls.
• Product snippet. The user cannot buy from this page (editorial review, comparison, third-party listicle). Lighter requirements; emphasis on Review, pros/cons, and aggregateRating rather than transactional fields.

AI shopping assistants extract from both. Editorial product-snippet pages are heavily cited for "best [category]" queries; merchant-listing pages win the buy-now intent.

Required and Recommended Fields

Based on schema.org Product and Offer:

For Offer:

GTIN, MPN, and SKU Identifiers

GTIN is the preferred unique identifier for global commerce (schema.org/gtin, GS1 GTIN reference). The gtin property generalizes the older gtin8/gtin12/gtin13/gtin14 and accepts both the bare numeric string and a GS1 Digital Link URL. Rules:

• Numeric: 8, 12, 13, or 14 digits.
• Must have a valid GS1 check digit.
• No left-padding; do not pad an UPC-12 to 13 with a leading zero outside formal conversion contexts.
• Use mpn when no GTIN is assigned (often DTC private label).
• sku is internal and does not replace gtin/mpn for AI shopping eligibility.

For pages selling third-party products, mirror the manufacturer GTIN, not your own SKU.

Minimum Compliant Example

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Acme Standing Desk 60\" Walnut",
  "image": ["https://example.com/desk-hero.jpg"],
  "description": "Electric height-adjustable standing desk with 60-inch walnut top.",
  "brand": { "@type": "Brand", "name": "Acme" },
  "sku": "ACME-DESK-60-WAL",
  "gtin": "0860001234567",
  "offers": {
    "@type": "Offer",
    "url": "https://example.com/p/acme-desk-60-walnut",
    "price": "599.00",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "priceValidUntil": "2026-12-31"
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.6",
    "reviewCount": "328",
    "bestRating": "5",
    "worstRating": "1"
  }
}

Multi-Variant Pattern with ProductGroup

For catalogs where one parent product has many SKUs (sizes, colors), use schema.org ProductGroup with hasVariant.

{
  "@context": "https://schema.org",
  "@type": "ProductGroup",
  "name": "Acme Standing Desk",
  "description": "Electric height-adjustable standing desk.",
  "productGroupID": "ACME-DESK",
  "variesBy": ["https://schema.org/color", "https://schema.org/width"],
  "hasVariant": [
    {
      "@type": "Product",
      "name": "Acme Standing Desk 60\" Walnut",
      "sku": "ACME-DESK-60-WAL",
      "gtin": "0860001234567",
      "color": "Walnut",
      "width": { "@type": "QuantitativeValue", "value": "60", "unitCode": "INH" },
      "offers": { "@type": "Offer", "price": "599.00", "priceCurrency": "USD", "availability": "https://schema.org/InStock" }
    },
    {
      "@type": "Product",
      "name": "Acme Standing Desk 72\" Walnut",
      "sku": "ACME-DESK-72-WAL",
      "gtin": "0860001234574",
      "color": "Walnut",
      "width": { "@type": "QuantitativeValue", "value": "72", "unitCode": "INH" },
      "offers": { "@type": "Offer", "price": "699.00", "priceCurrency": "USD", "availability": "https://schema.org/InStock" }
    }
  ]
}

Reviews and AggregateRating

• aggregateRating.ratingValue must match the visible average on the page.
• aggregateRating.reviewCount must match the visible review count.
• Use only first-party reviews you collected with disclosure; do not import third-party aggregations into aggregateRating.
• Include 2-5 individual Review items so AI engines can extract specific quotes.
• Review.author should be a Person with a real name, not a generic placeholder.

Adding aggregateRating for a value not visible on the page is a Google AI features compliance violation and a frequent cause of rich-result loss.

Google Merchant Center Alignment (2026)

The 2026 Merchant Center product data specification update added product-level shipping controls, a new [video_link] attribute (quality checks begin June 30, 2026), and image-resolution warnings ahead of January 31, 2027 enforcement (Google Merchant Center 2026 update help center). Recommendations:

• Submit the same product data via Merchant Center feed and on-page schema; mismatch flags Quality Issues that propagate to AI Overviews shopping eligibility.
• Add product video to both [video_link] feed attribute and a VideoObject reference on the product page.
• Render images at minimum 500x500 today; plan for higher minimum by Jan 2027.
• Reflect handling cutoff time and minimum order value at the product level if you submit them in the feed.

Five Real-World Examples

1. DTC apparel. ProductGroup parent + hasVariant per size/color; audience carries sizing system; first-party reviews drive aggregateRating.
2. Single-SKU specialty product. Flat Product with full Offer, aggregateRating, and MerchantReturnPolicy.
3. Editorial "best of" comparison. Product markup per featured product (product-snippet variant) without merchant-listing fields; emphasize review and aggregateRating for each.
4. Subscription box plan page. Product with subscription Offer (priceSpecification, eligibleQuantity); separate plan tiers as hasVariant on a ProductGroup.
5. Marketplace listing. AggregateOffer summarizing many seller offers; individual Offers in offers[] with seller Organization references.

Validator Quirks

• Offer.price as string vs number. Both validate; use string when including decimals ("599.00") to avoid floating-point representation surprises.
• availability URL form. Use full URLs like https://schema.org/InStock, not bare strings. Some validators tolerate the bare form; AI extractors are inconsistent.
• Multiple gtin types in same record. Use the generic gtin field; do not also set gtin13/gtin14.
• Missing brand. Validators warn rather than fail; AI engines down-rank brand-less products.
• Schema for products that are out of stock or not yet released. Set availability accurately (OutOfStock, PreOrder, BackOrder); do not omit the offer.
• Mismatch with on-page price. Hard fail in Google's pipeline; use server-side rendering of price in both schema and visible HTML.

Common Mistakes

• Importing third-party reviews into aggregateRating without disclosure.
• Padding GTIN with leading zeros incorrectly (UPC-12 to GTIN-13 rules vary; verify check digit).
• Using sku as the unique identifier when GTIN is available.
• Listing variants as separate top-level Product records instead of ProductGroup hasVariant.
• Forgetting priceValidUntil; AI engines treat unbounded prices as stale.
• Schema present but visible price hidden behind JS hydration that delays beyond the AI crawler render budget. Server-render the price.
• Putting Product schema on category pages without ItemList; Google has explicitly required ItemList wrapping for category-level Product markup historically.

FAQ

Q: Do I need both Product schema and a Merchant Center feed?

For AI shopping citations across ChatGPT, Gemini, and Google AI Overviews, both. The feed handles paid Shopping eligibility and Merchant Center quality validation; the page-level schema handles organic AI extraction. Keep them consistent.

Q: Should editorial review pages use merchant-listing or product-snippet markup?

Product-snippet. Merchant-listing requires fields that imply user can purchase on this page; misusing it on editorial pages produces validation issues and disqualifies the page from rich results.

Q: How does ChatGPT Shopping specifically use Product schema?

ChatGPT Shopping crawls with OAI-SearchBot and consumes Product schema as one of its selection factors alongside on-page text, reviews, and price (WebFX ChatGPT Shopping guide, Practical Ecommerce on ChatGPT Shopping schema). Detailed multi-facet schema increases recommendation likelihood.

Q: Do I need GTIN if I sell private-label DTC?

Use mpn instead of gtin for private-label products that lack a GS1 GTIN. Either is acceptable as the unique identifier requirement.

Q: How do I handle subscription pricing?

Use a Product with an Offer whose priceSpecification is a UnitPriceSpecification with unitCode and referenceQuantity. Provide separate Offers for each cadence (monthly, annual) or use hasVariant on a ProductGroup.

Q: What about used or refurbished items?

Set Offer.itemCondition to one of NewCondition, RefurbishedCondition, UsedCondition, or DamagedCondition. AI shopping assistants surface condition in answers.

Q: Do I include both first-party and review-platform aggregate ratings?

Use first-party only in aggregateRating for the product. Reference review-platform aggregates separately under subjectOf or in body text; merging them blurs attribution and creates compliance issues.

Q: What should I do for out-of-stock products?

Keep the Product and Offer markup with availability: OutOfStock. Removing the page or markup confuses AI extractors. If discontinued permanently, use availability: Discontinued.