JSON-LD Best Practices for AI Discoverability
Practical guidance on implementing Schema.org structured data that AI crawlers and knowledge graphs actually consume.
JSON-LD (JavaScript Object Notation for Linked Data) is the recommended format for embedding structured data in web pages. It's what Google, Bing, and AI systems prefer for consuming entity information. Done right, it makes your business machine-readable. Done poorly — or not at all — it leaves AI systems guessing.
This guide covers practical JSON-LD implementation for AI discoverability, based on what actually works in production.
Why JSON-LD Over Other Formats
Schema.org supports three syntaxes: Microdata, RDFa, and JSON-LD. Use JSON-LD. Here's why:
- Separation of concerns — JSON-LD lives in a
<script>tag in the<head>, completely separate from your HTML content. You can update structured data without touching your templates. - Google's explicit preference — Google recommends JSON-LD over Microdata and RDFa. Their documentation, tools, and examples all default to JSON-LD.
- AI consumption — AI crawlers parse JSON-LD more reliably than inline markup. It's a clean data format that doesn't require DOM traversal.
- Framework compatibility — JSON-LD works seamlessly with React, Next.js, Vue, and other modern frameworks. Microdata requires modifying rendered HTML elements.
Essential Properties
Every entity's JSON-LD should include these baseline properties:
{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Your Business Name",
"description": "A detailed description (200+ chars)",
"url": "https://yourdomain.com",
"telephone": "+1-555-000-0000",
"email": "contact@yourdomain.com",
"address": {
"@type": "PostalAddress",
"streetAddress": "123 Main Street",
"addressLocality": "Denver",
"addressRegion": "CO",
"postalCode": "80202",
"addressCountry": "US"
},
"sameAs": [
"https://linkedin.com/company/yourbusiness",
"https://twitter.com/yourbusiness"
]
}Choosing the Right @type
Schema.org has hundreds of entity types. Choose the most specific one that applies:
- Generic businesses:
LocalBusinessorOrganization - Professional services:
ProfessionalService,LegalService,AccountingService - Food & hospitality:
Restaurant,LodgingBusiness,FoodEstablishment - Healthcare:
MedicalBusiness,Dentist,Physician - Education:
EducationalOrganization,School,CollegeOrUniversity
More specific types give AI systems better context for categorization and recommendations.
Rich Properties That Matter
Beyond the basics, these properties significantly improve AI discoverability:
- openingHoursSpecification — Machine-readable operating hours. AI assistants use this to answer "Is X open right now?"
- areaServed — Define your service area. Critical for location-based AI recommendations.
- hasOfferCatalog — List your services with descriptions. Helps AI match your business to specific queries.
- image — A high-quality image URL. Used by AI systems for visual context and rich results.
- foundingDate — Signals business longevity and stability.
- aggregateRating — If you have reviews, include them. AI systems weight social proof heavily.
Common Mistakes
- Using Microdata instead of JSON-LD — Modern AI crawlers handle JSON-LD better. Switch.
- Minimal description — "We're a great business" tells AI nothing. Write 200+ characters describing what you do, who you serve, and where you operate.
- Missing sameAs — Every social profile you don't link is a missed cross-reference signal.
- Outdated data — Stale structured data (wrong address, old phone number) degrades trust faster than no data at all.
- Duplicate or conflicting markup — Multiple JSON-LD blocks with conflicting information confuse parsers. Use a single, authoritative block.
Testing Your Implementation
Before deploying, validate with:
- Google Rich Results Test — search.google.com/test/rich-results
- Schema.org Validator — validator.schema.org
- Nordax AI Visibility Score — Create a profile on Nordax AI to get a comprehensive score across all five dimensions of AI discoverability.
The Nordax Approach
Nordax AI generates JSON-LD automatically from your entity profile. When you fill out your profile, we output fully valid, comprehensive Schema.org markup that covers all the properties above — plus verification metadata. You get a hosted JSON-LD endpoint, an embeddable badge, and a public entity page — all machine-readable, all verified, all maintained.
You can manually implement JSON-LD on your own site, and you should. But Nordax AI provides the verified, cross-referenced entity layer that sits above raw markup — the trust infrastructure that makes your structured data actually trusted.
Ready to build your entity profile?
Create a structured, machine-readable identity for your business in minutes. Free to start.
Get Started Free