Schema for AI search: a practical guide

Schema used to be an SEO detail. In AI search it is part of entity identity. A model trying to cite you needs to know three things: who you are, what you are saying, and whether you are trustworthy. Valid structured data answers all three, on every single page.

The @graph pattern

Always use a single JSON-LD block with a @graph array. Nodes reference each other by @id. This is what a clean Article page looks like.

{
  "@context": "https://schema.org",
  "@graph": [
    { "@type": "Organization", "@id": "https://www.seotopsecret.com/#organization", "name": "SEOTopSecret" },
    { "@type": "WebSite", "@id": "https://www.seotopsecret.com/#website", "publisher": { "@id": "https://www.seotopsecret.com/#organization" } },
    {
      "@type": "Article",
      "@id": "https://www.seotopsecret.com/blog/schema-for-ai-search-a-practical-guide#article",
      "headline": "Schema for AI search: a practical guide",
      "author": { "@type": "Person", "name": "Javier Dominguez" },
      "publisher": { "@id": "https://www.seotopsecret.com/#organization" },
      "inLanguage": "en"
    },
    { "@type": "BreadcrumbList", "itemListElement": [ /* ... */ ] },
    { "@type": "FAQPage", "mainEntity": [ /* ... */ ] }
  ]
}

Minimum viable schema per page type

Blog post / article

1. Organization (root identity)
2. WebSite (with publisher reference)
3. Article (headline, author, datePublished, dateModified, image, inLanguage)
4. BreadcrumbList
5. FAQPage (only if FAQs are present on page)

Home page

1. Organization
2. WebSite (with SearchAction)
3. WebPage
4. SoftwareApplication or Organization-aligned primary entity

Pricing page

1. Organization
2. Product (with Brand)
3. Multiple Offer nodes (one per tier)

Entity linking via sameAs

The sameAs property on Organization and Person is how AI engines stitch your identity across sources. Point it at LinkedIn, X, Crunchbase, Wikidata (if you are there), and your GitHub organization. More specificity beats more entries.

{
  "@type": "Person",
  "name": "Javier Dominguez",
  "url": "https://www.seotopsecret.com/about",
  "sameAs": [
    "https://www.linkedin.com/in/javierdomm/",
    "https://twitter.com/_seotopsecret"
  ]
}

Automating this correctly

Hand-writing JSON-LD is error-prone — one missing comma breaks the whole graph. Use a schema builder that validates against Schema.org, output one @graph per page server-side, and never inject critical nodes via JavaScript. The Schema Generator inside SEOTopSecret does this automatically, but the pattern is what matters — not the tool.