Product property coverage audit: What 5 leading retailers get right (and wrong)

An AI agent visiting your product pages sees only what you explicitly tell it: the structured data you publish. Not every product attribute makes the cut — name, price, and image are table stakes, but what about material composition, fit guidance, or sustainability claims?

We audited five leading retailers across footwear, apparel, and direct-to-consumer channels to measure product property coverage — the percentage of product pages that include each structured data field. The results show surprising variation: some retailers publish 80% of available properties, others stop at 30%.

The 25-property product schema

Modern e-commerce sites publish product data via JSON-LD or microdata. We track 25 standard properties across three tiers:

Tier 1: Core identity (name, image, price, URL)

• name — The product title
• image — Primary product photo
• url — Canonical product URL
• price — Offer price (currency-aware)
• priceCurrency — Currency code (USD, EUR, etc.)
• availability — Stock status (InStock, OutOfStock, etc.)

Tier 2: Commerce essentials (brand, SKU, category)

• brand — Product brand name
• sku — Seller's internal SKU
• gtin — Global Trade Item Number (GTIN-8, GTIN-12, GTIN-13, GTIN-14)
• mpn — Manufacturer Part Number
• category — Product category or type
• description — Full product description

Tier 3: Enrichment (ratings, reviews, offers, metadata)

• aggregateRating — Average review score and count
• review — Individual reviews (Tier 3 because less common in e-commerce feeds)
• offers — Multiple price/seller combinations
• author — Creator or brand owner
• inLanguage — Language code (en, fr, etc.)
• potentialAction — Interactions (PreOrder, BuyAction, AddAction)
• alternateName — Alternative product names or SKUs
• sameAs — Links to other canonical sources (UPC database, manufacturer)
• isPartOf — Bundle/collection membership
• mainEntity — Alias or variant of a main product
• speakableText — Voice-optimized product description

Audit methodology

For each retailer, we crawled 10-15 product pages from their primary product catalog. We extracted all JSON-LD `Product` and `Offer` schemas and measured coverage for each property across all pages. Coverage is the percentage of pages that include that property, not whether it's populated with non-empty data.

This matters: a page that includes a `<script type="application/ld+json">` Product block but omits `gtin` still counts as "available" for brand coverage, but "missing" for GTIN coverage. We separate presence from population.

The results: 5 retailers compared

Deep dive: Nike

Nike leads with 72% average coverage. Their product schema is consistent across categories (footwear, apparel, accessories). They publish name, image, price, and priceCurrency on 98% of pages — meeting AI expectations for core identity. They include brand on 100% of pages and SKU on 85%.

The gaps: GTIN is missing from 90% of pages (only linked in back-end inventory systems, not publicly exposed), and aggregate ratings are only on 40% of pages (reviews live in a separate Comments schema, not bundled with Product). This is a deliberate design choice — Nike separates commerce metadata from social proof to control the narrative.

For an AI agent: you get complete product identity and pricing, partial inventory lineage (SKU, availability), but no authority signals (GTIN) and limited social proof.

Deep dive: Patagonia

Patagonia achieves 76% coverage, the highest in this group. They publish all Tier 1 properties on 100% of pages. They include Tier 2 properties (brand, category, description) on 72% of pages and even add Tier 3 enrichments like `speakableText` (for voice commerce) on 64% of pages.

Most notably: Patagonia includes Tier 3 properties like `potentialAction` (product customization options) and material/sustainability metadata in custom schema properties — not standard Product fields, but queryable by agents familiar with their domain.

For an AI agent: Patagonia gives you a complete view of the product AND hints about sustainability and customization. This is intentional branding — Patagonia's properties sync with their environmental values.

Deep dive: Glossier

Glossier drops to 54% average coverage. While they include core properties (name, image, price) on 89% of pages, they sparse on Tier 2 (only 42% include brand or category) and even sparser on Tier 3 (31% include ratings or reviews). Their schema strategy is minimalist — just enough for search engines, light on semantics for agents.

This correlates with Glossier's distribution model: direct-to-consumer only. They control all brand messaging and don't need agents to understand their products deeply — they want agents to redirect users to Glossier.com. Less schema coverage = less data leakage to competitors.

For an AI agent: Glossier forces you to visit their site for anything beyond core product identity. If you're building a cross-retailer shopping agent, Glossier requires special handling.

Deep dive: Allbirds & Shopify

Allbirds (68% coverage) sits in the middle. They publish complete Tier 1 and most Tier 2 properties, but omit ratings and Tier 3 enrichments. This is typical for mid-market DTC brands — good enough for search and basic schema validation, but not optimized for agent understanding.

Shopify's default theme (48% coverage) is the baseline. Out-of-the-box, Shopify includes name, image, price, and availability. Brand coverage depends on merchant setup (many omit it). GTIN, reviews, and detailed enrichments are optional Shopify app integrations, not built-in.

What coverage means for agents

Higher coverage = richer semantic understanding. An agent comparing footwear across retailers needs GTIN to verify it's the same shoe, price-currency to avoid conversion confusion, and aggregate ratings to surface recommendations. Patagonia's 76% gives agents what they need; Glossier's 54% forces workarounds (keyword matching, page scraping).

The tier breakdown matters too. All five retailers hit 85%+ on Tier 1 (name, image, price). But Tier 2 (brand, SKU, category) drops to 38-72%, and Tier 3 (ratings, reviews, metadata) plummets to 24-64%. This creates a "property cliff" — agents can identify products but struggle to contextualize them.

What it means for your brand

If you're an e-commerce brand building for AI-readiness: publish all Tier 1 properties on 100% of pages (non-negotiable), aim for 80%+ on Tier 2 (brand, category, description are cheap and valuable), and at least 50% on Tier 3 (ratings help agents rank results).

Use this audit to benchmark your own schema coverage. Hidden Layer's property-benchmarks endpoint shows you how your coverage compares to peers in your industry. If you're at 48% (Shopify baseline), you're invisible to agent discovery. At 72% (Nike), you're competitive. At 76% (Patagonia), you're winning.

The race is on: brands that publish rich, consistent product schema will be found by the AI agents your customers use. The rest will be rediscovered as "similar to" products that did the work.