A schema checker built for LLM extraction, not just rich results.

What we validate

• Syntax: valid JSON-LD, no parser errors, no orphan nodes.
• Type correctness: required and recommended properties per Schema.org spec.
• Rich result eligibility: types Google explicitly grants rich treatment to.
• Entity disambiguation: sameAs edges to Wikidata, Wikipedia, LinkedIn, Crunchbase, GitHub.
• Author/Person schema depth: name, jobTitle, knowsAbout, credentials, sameAs.
• Organization schema completeness: logo, address, contactPoint, sameAs, founder.
• Article schema: headline, datePublished, dateModified, author, image, mainEntityOfPage.
• FAQPage and HowTo for passage extraction.
• BreadcrumbList for SERP and entity-path clarity.
• Aggregate signals: schema coverage % across the audited pages.

Why schema for LLM extraction is different

Google grants rich results to a curated list of types. LLMs read every type in your JSON-LD as evidence about the entities on the page. That means schema you write only for "rich results" misses the opportunity to bind your content to the entity graph LLMs use to disambiguate brands, products and authors. Two pages with identical content, where one has Person + Organization schema with sameAs edges and the other has none, get cited at very different rates in AI answers.

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "How AI Overviews choose citations",
  "datePublished": "2026-05-01",
  "dateModified": "2026-05-15",
  "author": {
    "@type": "Person",
    "name": "Jane Doe",
    "jobTitle": "Head of SEO Research",
    "knowsAbout": ["GEO", "AIO", "Structured Data"],
    "sameAs": ["https://www.linkedin.com/in/janedoe"]
  },
  "publisher": {
    "@type": "Organization",
    "name": "Acme",
    "sameAs": ["https://www.wikidata.org/wiki/Q123"]
  }
}

What we won't tell you to do

We won't recommend "speakable" schema for general content (it's specifically scoped to news), or stuff schema with fake aggregateRating (Google penalizes that), or add Review schema without real reviews. Invalid or spammy schema is worse than no schema. Findings include the fix and the rationale.

The 6-point schema checklist

1. Validate every JSON-LD block at parse time — orphan nodes kill the whole graph.
2. Ship Organization with logo, url, and 3+ sameAs edges.
3. Wrap every byline in Person with credentials and a LinkedIn sameAs.
4. Add Article with datePublished AND dateModified on every long-form page.
5. Use BreadcrumbList on category and product pages.
6. Never invent ratings, reviews, or events you don't have.