What Is JSON-LD?

In one sentence

JSON-LD is the recommended format for structured data: a block of JSON in a script tag describing what a page is about.

What JSON-LD means

It stands for JSON for Linked Data. Unlike older formats such as microdata, it sits separately from your visible HTML rather than being woven through it.

That separation is the main practical advantage — you can change your page design without breaking your markup, and vice versa.

Why it matters for AI search

For AI systems, JSON-LD provides unambiguous facts that don't have to be inferred from prose or reconstructed from scattered HTML attributes.

It's also far easier to generate dynamically from source data, which is what keeps it accurate as content changes.

See it in action

The same fact, three markup formats.

Why JSON-LD won

Format comparison
// click generate

Separation is the whole argument.

Markup woven through HTML breaks whenever someone changes the design. A separate JSON block generated from source data stays accurate.

How to get it right

Implementing JSON-LD

  • Place it in a script tag with type application/ld+json, in head or body
  • Generate values from source data rather than hardcoding them
  • Nest related entities using @id references so relationships are explicit
  • Validate after every release; a syntax error silently disables the whole block
  • Keep one coherent graph per page rather than several disconnected blocks

Common questions

Is JSON-LD better than microdata?

For most purposes yes. It's Google's recommended format, doesn't break when the design changes, and is far easier to generate programmatically.

Where should the script tag go?

Either head or body works. Head is conventional and keeps it away from content edits.

Can invalid JSON-LD hurt us?

A syntax error means the block is ignored entirely, wasting the effort. Markup that misrepresents page content can be treated as manipulative.

These come up alongside JSON-LD constantly.

Free 12-point check

Send Ron your site

Two fields. We’ll crawl your site as GPTBot, ClaudeBot and PerplexityBot, benchmark a sample of your category’s prompts, and send you what we find.

  • What each AI crawler actually receives from your site
  • Whether models resolve your brand as a real entity
  • A sample of category prompts and who gets cited
  • The three fixes we’d prioritise first

No commitment, no sales sequence. If we’re not a fit, we’ll say so.

We’ll pull your favicon so you know we found the right site.

We reply from a real address. No sequence, no newsletter.