In short:Schema.org Product in 2026: the 12 fields AI agents read (and the 3 Google ignores) — Since 2011, Schema.org Product has existed for Google. Rich Snippets, Google Shopping, Knowledge Graph.
18%of e-merchants with complete Schema Product (12 fields)
+67%AI citations with complete implementation
3critical fields ignored by most SEO plugins
Schema Product in 2026: a new reader
Since 2011, Schema.org Product has existed for Google. Rich Snippets, Google Shopping, Knowledge Graph.
In 2026, it has a new reader. More demanding. More precise. More influential on purchasing decisions.
AI shopping agents — ChatGPT Shopping, Perplexity Commerce, Google agents powered by Gemini — read Schema.org Product differently. They don’t rank pages. They answer purchase questions with precision.
This shift in purpose changes which fields matter.
18%of e-merchants with complete Schema Product (12 fields)
+67%AI citations with complete implementation
3critical fields ignored by most SEO plugins
The 12 fields read by AI agents
Here are the 12 Schema.org Product properties that structure visibility in AI shopping agents:
Identification fields
name — Exact product name. No marketing suffix ("Ultra PRO Coffee Maker"). Raw catalog name.
sku — Internal reference. So the agent can cross-match your sheet with third-party product databases.
gtin13 / gtin8 / mpn — Universal identifier. HIGH PRIORITY The agent checks this field to reconcile your product with other sources.
Commercial fields
offers.price — Numeric price, consistent with the visible page.
offers.priceCurrency — ISO 4217 code ("EUR", "USD"). Never the symbol alone.
offers.availability — Full Schema.org URL: https://schema.org/InStock
offers.priceValidUntil — ISO 8601 date. AI AGENT Agents verify this date for real-time promos.
Authority fields
brand.name — Exact brand name.
aggregateRating.ratingValue and aggregateRating.reviewCount — Rating and review volume. Trust signal.
description — 120-300 characters. Concrete benefits and measurable characteristics.
category or additionalProperty — Classification in product taxonomy. AI AGENT
The 3 fields Google ignored
Google values Schema Product for Rich Snippets. Stars, price, availability in the SERP. Three fields remained in the background. AI shopping agents read them all the time.
1. gtin13 / mpn — The universal identifier
Google used it for Shopping (Merchant Center feed). In organic search, no visible impact.
For an AI shopping agent, it's a trust pivot. With the GTIN or MPN, the agent cross-matches your sheet with the manufacturer database, third-party reviews, competitor prices. No universal identifier? A black box the agent can't validate.
Result: products with GTIN receive 2.3× more AI recommendations than those without, same category.
2. additionalProperty — Technical characteristics
Google rarely displayed it in organic search. AI shopping agents exploit it systematically for comparative questions.
"Which French press weighs less than 400g?" The agent reads additionalProperty for weight. Without this field, your product doesn't answer that specific question.
Google never really exploited this field for ranking. AI agents use it to evaluate the reliability of your prices.
A price with no validity date? Potentially obsolete. A price with a 30-day expiration? Verified. Current. This freshness signal directly influences the likelihood an agent will recommend you in their answer.
WooCommerce. Yoast and RankMath output 6-7 fields. For the full 12, install "Schema & Structured Data for WP & AMP" or use a custom woocommerce_structured_data_product filter.
Shopify. The native theme delivers basic Schema Product. To enrich it: JSON-LD for SEO, Schema Plus for SEO, or inject JSON-LD into product.json.liquid via Liquid snippet.
Validation required: After every Schema change, double-check both search.google.com/test/rich-results AND validator.schema.org. Both tools have different rules — validating for Google is not enough. AI agents expect full Schema.org compliance.
Data as competitive advantage
Schema Product is no longer a technical SEO exercise. It's the interface between your catalog and the AI agent economy.
Merchants who master these 12 fields today are building a data infrastructure their competitors will have to replicate — from scratch, when the market is already structured.
The GTIN. The priceValidUntil. The additionalProperty. Three fields. A few hours of implementation. A structural advantage on AI citations that few merchants exploit yet.
Technical JSON-LD implementation: complete examples for 3 product types
Schema theory is well documented. Complete examples for real-world e-commerce are less so. Here are three tested and validated JSON-LD implementations covering the most common cases.
Bundles. Nearly invisible in Schema.org even though AI agents love them. Use isRelatedTo to link components. And description to list exactly what the pack contains. No mystery.
Validate every JSON-LD with Google's Rich Results Test before pushing it to production. Malformed JSON won't trigger any visible error — it will simply be ignored. Discovering the error three months later, after losing traffic, is expensive.
Emerging fields in 2026: hasEnergyConsumptionDetails, isFamilyFriendly, additionalProperty
Schema.org is a living standard. In 2026, three fields are gaining weight because AI agents are integrating sustainability and personalization into their recommendations.
hasEnergyConsumptionDetails — for appliances and electronics
This field structures energy class according to EU standards. Since ChatGPT and Perplexity are integrating environmental criteria, products with this field see their citation rate climb on "economic", "consumption", "low-consumption" queries.
isFamilyFriendly — for merchants targeting families
Simple boolean. Documented impact on parental queries. Shopping agents include it as soon as the query mentions children, babies, family context. Value: true or false. Never empty — an explicit value beats missing information.
additionalProperty — the most powerful and under-used field
additionalProperty structures any product characteristic not covered by standard Schema properties. This is where you gain the most ground on competitors.
+28% citations on specialized queries for products using additionalProperty with 5+ structured characteristics — Q1 2026 analysis
Testing and validating your Schema Product: tools and protocol
Deploying Schema without validation? Most common error. Syntactically correct JSON-LD can be semantically wrong — and ignored by agents. Complete protocol.
Step 1 — Syntax validation
Tool: Rich Results Test (search.google.com/test/rich-results). Paste your URL or JSON-LD code. The test returns syntax errors and missing properties for each eligible rich result type.
Complementary tool: Schema Markup Validator (validator.schema.org). More strict. Verifies full compliance with Schema.org specs.
Step 2 — Semantic validation
Syntax confirms your JSON is correct. Semantics confirms values match reality.
Semantic checklist:
Price matches displayed page price — zero tolerance
brand.name matches your official brand name exactly
Step 3 — Continuous monitoring
Search Console → Enhancements → displays Schema errors detected across your site. Set up email alerts for every new Schema error. A healthy e-commerce maintains zero critical errors and fewer than 50 warnings across 10,000 pages.
Schema errors that block AI shopping agents
Some Schema errors trigger no warnings in validation tools. They silently block agent reading. In silence.
Error 1 — Price without priceCurrency. Price without currency is ignored by 100% of AI shopping agents. No assumption. Without priceCurrency: "EUR", your price doesn't exist for them.
Error 2 — Availability with non-standard value. Acceptable values are only Schema.org URLs (https://schema.org/InStock, etc.). Writing "In stock" in English creates a silent error.
Error 3 — Image with relative URL. Image URL must be absolute (https://...). A relative URL like /images/product.jpg is invalid in JSON-LD.
Error 4 — Price with European decimal format. JSON-LD expects a decimal number with period: 29.90. Writing 29,90 breaks parsing in most agents.
Error 5 — AggregateRating with reviewCount of 0. A product with 0 reviews and an AggregateRating schema is interpreted as manipulation. You have real reviews, you declare them. You don't have any, you don't include the AggregateRating block.
Quick audit: take your 10 most important product sheets. Paste their URLs in Rich Results Test. If more than 3 have critical errors, that's your priority #1 before any other GEO or SEO optimization.
Prioritizing Schema implementation: the impact matrix
You have 10,000 product sheets. Where to start? The impact matrix lets you prioritize the 20% of effort that generates 80% of gains.
The 4 priority quadrants
Quadrant 1 — Maximum impact, low effort. Add price, availability, and priceCurrency to all sheets missing them. These 3 fields deploy via a global template in a few hours. Impact: +35 to +50% presence in ChatGPT Shopping depending on sector.
Quadrant 2 — High impact, moderate effort. Add AggregateRating to your 200 bestsellers. Requires aggregating existing reviews in the markup. 2 to 3 days of development for a WooCommerce store with the right plugin. Impact: products with well-filled AggregateRating are 2.8x more likely to be recommended on comparative queries.
Quadrant 3 — High impact, significant effort. Implement ProductGroup + variants for your 50 top sellers in multiple colors/sizes. Effort: 5 to 8 days of development. Impact on specialized queries (color, size, style): +67% presence measured.
Quadrant 4 — Moderate impact, variable effort. Emerging fields (hasEnergyConsumptionDetails, additionalProperty, certification). Deploy in wave 3, on products where these attributes are relevant.
Wave deployment: 90-day plan
Days 1 to 15: Complete Schema audit via Rich Results Test on 100 representative URLs. Map critical errors and opportunities by product type.
Days 16 to 45: Deploy Quadrant 1 corrections across the entire catalog. Validate via Search Console Enhancements.
Days 46 to 75: Deploy Quadrant 2 on top 200 bestsellers. Measure impact on ChatGPT Shopping traffic.
Days 76 to 90: Deploy Quadrant 3 on 50 variable references. Begin emerging fields on relevant catégories.
Measuring Schema ROI
Three metrics directly attributable to Schema improvement:
Rich Results display rate in Search Console (before/after for each wave)
Clicks via ChatGPT Shopping in your analytics referral (source: openai.com or chatgpt.com)
Perplexity citation rate on 30 benchmark product queries (monthly measurement)
Schema is not optional — it's the interface between your catalog and AI agents. In 2026, a catalog without complete Schema Product becomes invisible to 30% of emerging e-commerce traffic. Investment pays back in 60 to 90 days on well-optimized catégories.
Frequently asked questions
Should every product sheet have Schema Product, or just the best ones?
Every product sheet — but with completeness adapted to sheet importance. For your 100 top sellers: full 12 fields including additionalProperty and GTIN. For less strategic sheets: 7 core fields (name, sku, brand, price, priceCurrency, availability, image). Incomplete Schema on a secondary sheet beats no Schema.
How do I get GTINs for my products?
Three sources: 1/ Manufacturers provide GTINs in technical sheets or distributor catalogs. 2/ GS1 database (gs1.org/services/verified-by-gs1) lets you verify existing codes. 3/ For your own references, assign your own EAN codes via GS1 France (gs1fr.org). For imported products, GTINs are often in the supplier's PDF catalog or via EDI access.
Does Schema Product duplicate Google Merchant Center feed?
No. Two separate channels. Merchant Center uses a separate CSV or XML feed for Google Shopping paid and free. Schema.org Product is read by Google for Rich Snippets and by AI agents for recommendations. Data must be consistent between both, but both channels are complementary, not redundant.
What happens if Schema price differs from displayed price?
Two consequences. For Google: Rich Snippets penalty risk (warning in Search Console, then loss of stars and price in results). For AI agents: trust loss and temporary recommendation exclusion. AI agents test data consistency more aggressively than Google. More than 2% variance typically triggers exclusion. Solution: automate Schema sync from your product database on every price update.
Does Schema Product work for variable products (sizes, colors)?
Yes, with hasVariant property. Each variant (size S/M/L, color red/blue) can be declared as child Product with its own offers fields (price, availability, SKU). On WooCommerce, the right Schema plugin handles variants automatically. On Shopify, native variant schema is limited — a dedicated app is recommended for catalogs with many variants.
Schema Product audit of your catalog
I test your current Schema Product live, identify the 3 critical missing fields, and give you an implementation plan tailored to your CMS.
