What Is JSON-LD and Why It Matters for LLM Visibility

When ChatGPT, Claude, or Perplexity crawls your website, they do not read it like a human. They scan the HTML looking for structured signals — machine-readable hints about what the page is, who runs it, and what it covers. The most important of those signals is JSON-LD.

What JSON-LD actually is

JSON-LD (JavaScript Object Notation for Linked Data) is a format for embedding structured data in your HTML. It uses the vocabulary from schema.org, the shared standard maintained by Google, Microsoft, Yahoo, and Yandex since 2011.

In practice, JSON-LD is a <script> tag you drop inside the <head> of your page:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Acme",
  "url": "https://acme.com"
}
</script>

That is it. One script tag per schema, sitting inside the head, invisible to users but critical for crawlers.

Why LLMs care about it

Without JSON-LD, an LLM has to guess what your page is about by parsing raw HTML. With it, the model reads a precise, typed description: "this is an Organization named Acme, its URL is X, its logo is Y, and it is the publisher of this WebSite."

That difference compounds. A model citing your business by name needs to know your name. A model recommending a FAQ answer needs to know which questions you actually answer. JSON-LD gives that information in the cleanest form possible.

The four schemas every site should have

You do not need every schema type. For 90% of sites, these four cover the essentials:

1. Organization

Describes the company or brand behind the site — name, URL, logo, description. This is the schema that lets an LLM say "Acme is a fintech founded in Mexico" instead of guessing.

2. WebSite

Describes the site itself — name, URL, language. It sounds redundant with Organization, but it is a different schema type that some crawlers and search engines explicitly look for.

3. BreadcrumbList

Shows navigation hierarchy. Even a homepage-only breadcrumb (just a single "Home" entry) is enough to trigger the detection in most crawlers, including LLM-focused ones.

4. FAQPage

Turns your FAQ section into citable question-answer pairs. If a user asks ChatGPT a question that matches one of your FAQ items, the model can cite your answer directly. This is one of the highest-leverage schemas you can add.

Where to paste the scripts

All four go inside the <head> of the relevant page:

Organization and WebSite: every page (best added in your site layout)
BreadcrumbList: every page (regenerated per page with the correct hierarchy)
FAQPage: only on pages that actually have an FAQ section visible to users

You can either include multiple schemas in a single script using @graph, or keep them in separate <script> tags. Both work. Google's Rich Results Test is a fast way to verify that your markup parses correctly.

Common mistakes

Lying to the schema. If your site does not actually have the FAQ listed in FAQPage, crawlers notice and downgrade your trust. Match schema to reality.
Copy-pasting without customizing. A generic Organization schema with "name": "Your Company" does nothing. Fill in your real values.
Forgetting the logo. Organization without a logo URL is weaker — the logo is how LLMs visually identify your brand in answer cards.
Relative URLs. All URLs in schemas must be absolute (https://...), never relative (/about).

The lazy way to get all four

If writing JSON-LD from scratch feels like homework, run your site through LLCrawler. The report includes a JSON-LD kit built from your actual site data — Organization with your real name and logo, WebSite with your language, BreadcrumbList for your homepage, and FAQPage if we detected a FAQ section. Copy each block, paste into your <head>, done.

JSON-LD is not optional infrastructure for sites that want to be cited by AI. It is the basic vocabulary. Without it, you are invisible in the conversation.