Structured data: an agent's native language

Why it's every agent's first move

When an agent lands on a product page, it doesn't “look” at your layout: it looks first for schema.org markup. That's the machine version of the page, the one feeding comparisons — and, for years now, Google's rich results. Complete, accurate markup means being understood; missing or wrong markup means asking the agent to guess. It doesn't guess: it moves to a competitor.

The core: Product, Offer, Organization

Three types cover the essentials of a merchant site:

• Product — the item: name, reference (sku, gtin), brand, description, attributes.
• Offer — the commercial offer: price, currency, availability, delivery and return terms. The part agents scrutinise most.
• Organization — the merchant: identity, contact details, customer service. The raw material of trust.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "18V Pro cordless drill",
  "sku": "PV18-PRO",
  "brand": {"@type": "Brand", "name": "Fabrix"},
  "offers": {
    "@type": "Offer",
    "price": 89.00,
    "priceCurrency": "EUR",
    "availability": "https://schema.org/InStock",
    "priceValidUntil": "2026-12-31",
    "eligibleQuantity": {"@type": "QuantitativeValue", "minValue": 6}
  }
}

Extend it to your content: FAQPage on frequently asked questions, Article on the blog, BreadcrumbList everywhere, and DefinedTerm if you publish a glossary — generative engines love those.

The five costliest mistakes

1. Markup price differs from displayed price. The inconsistency is detected, and your whole credibility is downgraded.
2. Availability never updates. Permanent InStock over real stockouts means agentic transactions will fail.
3. Marking up only part of the catalog. To an agent, unmarked products are second-class products.
4. Ignoring B2B fields. Minimum quantity (eligibleQuantity), net prices, packaging: without them your professional offer is unreadable.
5. Marking up wrong data. Markup amplifies what it structures — errors included. Repository clean-up always comes first.

Method and validation

1. Generate the JSON-LD dynamically from your product database — never by hand, never hard-coded.
2. Validate with Google's Rich Results Test and the schema.org validator, on a sample of every page template.
3. Monitor in Search Console: markup errors surface there continuously.
4. Audit consistency across markup ↔ page ↔ feed: all three must tell exactly the same story.