Structured Data (JSON-LD): Definition for AI Search

JSON-LD structured data is semantic metadata placed inside a <script type="application/ld+json"> tag in a web page's source code, written in the standardized Schema.org vocabulary. It lets search engines and AI crawlers read a page's key facts directly — content type, author, date, topic, related entities — without interpreting the visual rendering or the way the text is laid out.

Why it matters for AI visibility

JSON-LD is the main channel through which LLMs read a page's structured metadata. Unlike microformats or HTML meta tags, JSON-LD is a self-contained block of data that can be parsed on its own, independently of the DOM, with no JavaScript to run. AI crawlers that don't render JavaScript can read it right out of the raw HTML.

For AI visibility, a few JSON-LD properties matter more than the rest:

• Article or BlogPosting with headline, description, and mainEntityOfPage: the baseline signal that the page is a real article, not a product page or a navigation page.
• dateModified: a direct freshness signal. LLMs favor recent content, and an up-to-date dateModified shows the page is being maintained. This field should be present and equal to or later than datePublished.
• author with sameAs: the author should have a verifiable profile (LinkedIn, an institutional page) referenced through the sameAs property. This is E-E-A-T made concrete inside your structured data — an anonymous or unverifiable author runs up a credibility deficit.
• @graph: linking several entities together (Article, Author, Organization) into a connected graph is what LightSpot scores as criterion G17. Connected entities strengthen the model's confidence that the metadata is coherent.

LightSpot scores five structured-data criteria (G13 through G17), because they form the most direct line of communication between your content and the LLMs.

A concrete example

Here's a minimal Article JSON-LD block, tuned for LLMs:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "How to Choose CRM Software in 2026",
  "description": "A side-by-side guide to the 8 leading CRMs for small businesses, with a scoring grid and real-world feedback.",
  "datePublished": "2026-01-15",
  "dateModified": "2026-05-20",
  "author": {
    "@type": "Person",
    "name": "Marie Dupont",
    "jobTitle": "CRM Consultant",
    "sameAs": "https://www.linkedin.com/in/marie-dupont-crm"
  },
  "publisher": {
    "@type": "Organization",
    "name": "Acme Advisory",
    "sameAs": "https://www.linkedin.com/company/acme-advisory"
  }
}

This schema tells the LLM: this is an article (not a product page), written by an identifiable and verifiable person, kept up to date recently, and published by an organization with a public profile. Each of those facts feeds the trust the model places in the passage when it's picking sources to cite.

For the full picture, see the complete GEO guide.