Product schema markup for AI shopping agents: the WooCommerce guide

One mechanism, two eras

This guide assumes you already know the basics of schema markup — if not, start with our complete guide to schema markup for AI visibility. Here, we go deeper on a single type : Product, applied to WooCommerce.

The principle hasn't changed since Google Shopping : any engine that needs to display a price, availability, and an average rating needs to read those somewhere structured, not guess them from an HTML page. What's changed is how many systems now read that schema.

What Google already requires — and hasn't changed

Google Merchant Center has spelled out its rules for years : a product identifier (GTIN) is required for most products ; where none exists, MPN combined with brand serves as a fallback ; for a genuinely unique product (handmade, custom-made), the identifier_exists: false attribute avoids inventing an identifier that doesn't exist. A product with a missing or incorrect GTIN gets limited visibility in Merchant Center — Google says so explicitly in its own documentation.

What's changing with AI shopping agents

Two mechanisms coexist here, and they shouldn't be conflated.

The dedicated merchant feed. ChatGPT's Instant Checkout (the protocol OpenAI documents as the Agentic Commerce Protocol) works through a structured product feed (CSV or JSON) that the merchant sends directly to OpenAI — not by reading the schema.org markup on your pages. OpenAI's own developer documentation is clear : required fields ensure price and availability display correctly ; recommended fields (reviews, rich media, performance signals) improve ranking and trust. Perplexity's merchant program follows a similar logic : a product catalog submitted separately, in a format modeled on Google Shopping's.

Ordinary page reading. Outside those dedicated feeds, an AI agent browsing the web normally (a ChatGPT or Perplexity answer in regular conversation mode, not the built-in checkout flow) reads your pages the way any visitor would — and clean Product/Offer schema lets it extract the correct price and availability instead of guessing from your page layout.

What ties the old world (Google Shopping) to the new one (AI merchant feeds) together : the GTIN remains the pivot identifier used to match your product to the right listing, regardless of which system is reading it.

The problem with WooCommerce's native schema

WooCommerce added a native global identifier field (GTIN, UPC, EAN, or ISBN) starting in version 9.2 — it syncs to Google Merchant Center through the Google for WooCommerce extension. That's a real improvement over before, when only third-party plugins or custom fields covered this need.

But two blind spots remain : WooCommerce still has no native field for MPN (a feature request remains open on its own GitHub repository) ; and the default schema.org output has long-documented gaps — brand, ratings, and identifiers aren't always complete depending on your theme and installed extensions. Never assume your theme handles this markup correctly without checking it yourself, at least once.

Checklist — check your own catalog

• Pick 3 to 5 real product pages and run them through search.google.com/test/rich-results.
• Check for GTIN, MPN, and brand — and confirm the GTIN is exactly 8, 12, 13, or 14 digits, or it gets ignored.
• Check that the price and availability in the schema exactly match what's shown on the page (and your merchant feed, if you submit one).
• Check that every product variation exposes its own Offer — not just the parent product.
• Check that AggregateRating reflects only real customer reviews — never an invented or inflated rating.

What Livada SEO does here — and what it doesn't

Livada SEO automatically emits Product, Offer, AggregateRating, and Review schema for WooCommerce, including product variations. That covers the on-page markup part of this guide — not submitting a feed to OpenAI's or Perplexity's merchant programs, which stays a separate step if you want to appear there, nor entering your GTIN/MPN values themselves, which remain your product data.

A clean checklist and correctly emitted schema don't replace each other : one tells you where you stand, the other applies it automatically once you know what to check.