Schema markup for AI agents: the complete guide

Why structured data matters for AI — not just Google

Traditional SEO treats structured data as an enhancement. It helps Google show star ratings, pricing, and FAQs in search results. Useful, but optional — the underlying page still ranks based on content, links, and technical signals. For AI agents, structured data is not optional. It is the foundation of understanding.

When an AI agent in the B2A2C model researches businesses on behalf of a user, it needs to extract precise, comparable facts: What services do you offer? What are the prices? Where are you located? What do customers say? Human visitors infer these details from context, layout, and design. AI agents cannot — they rely on explicitly structured data to parse this information reliably.

The data supports this shift:

• Websites with comprehensive Schema markup are 40% more likely to be cited in AI-generated answers compared to pages with equivalent content but no structured data, based on analysis of Perplexity and ChatGPT citation patterns.
• Only 33% of websites implement any Schema.org markup at all, and far fewer do so comprehensively — creating a significant competitive gap.
• AI agents process structured data 10 to 100 times faster than unstructured HTML, making schema-rich pages more likely to be selected when agents operate under time or token constraints.
• 76% of AI Overview citations come from pages in the Google organic top 10 — but within that set, pages with richer structured data are favored for citation.

How AI agents parse structured data differently from Google

The critical difference: Google uses structured data for display, while AI agents use it for decisions. An incomplete Schema implementation might cost you a rich snippet on Google. An incomplete Schema in the age of AI agents might cost you a recommendation to a paying customer.

Key Schema.org types for AI visibility

Not all Schema types carry equal weight for AI agents. The following types form the core of an AI-optimized structured data strategy:

Organization and ProfessionalService

This is the foundation for any business. It tells AI agents who you are, where you operate, and how to contact you. Use ProfessionalService for agencies, consultancies, and professional firms. Use LocalBusiness (or more specific subtypes like Restaurant or MedicalClinic) for location-dependent businesses. The Organization type works as a broader container for companies that do not fit neatly into a local category.

{
"@context": "https://schema.org",
"@type": "ProfessionalService",
"@id": "https://example.com/#organization",
"name": "Example Agency",
"description": "Digital agency specializing in web development and AI visibility for B2B companies.",
"url": "https://example.com",
"areaServed": ["Belgium", "France", "Luxembourg"],
"address": {
  "@type": "PostalAddress",
  "addressLocality": "Brussels",
  "addressCountry": "BE"
},
"priceRange": "EUR EUR EUR",
"knowsLanguage": ["fr", "en", "nl"],
"foundingDate": "2024",
"numberOfEmployees": {
  "@type": "QuantitativeValue",
  "minValue": 1,
  "maxValue": 10
}
}

Service and Offer

These types describe what you sell and at what price. AI agents performing comparisons rely heavily on Service and Offer to match user requirements against provider capabilities. Every distinct service should have its own Service entity with associated Offer and PriceSpecification.

{
"@type": "Service",
"name": "Website Design & Development",
"description": "Custom, high-performance website built with modern frameworks, optimized for conversion and AI visibility.",
"provider": { "@id": "https://example.com/#organization" },
"areaServed": "Belgium",
"offers": {
  "@type": "Offer",
  "priceSpecification": {
    "@type": "PriceSpecification",
    "priceCurrency": "EUR",
    "minPrice": "3000",
    "description": "Starting price for showcase websites"
  }
}
}

Product

For e-commerce businesses, Product schema with detailed attributes — SKU, availability, price, brand, reviews — is essential. AI shopping agents are already comparing products across stores in real time. Without complete Product schema, your items are invisible to these agents. Include AggregateRating and Review entities wherever possible, as social proof signals strongly influence AI agent recommendations.

FAQPage

FAQ markup is one of the most effective Schema types for AI visibility. AI agents routinely extract FAQ data to answer user questions directly, and pages with FAQPage schema are significantly more likely to be cited in AI-generated responses. Each question-answer pair becomes a discrete, citable fact.

{
"@type": "FAQPage",
"mainEntity": [
  {
    "@type": "Question",
    "name": "How much does a website cost in Belgium?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Website costs in Belgium range from EUR 3,000 for a showcase site to EUR 30,000+ for complex e-commerce or marketplace platforms."
    }
  },
  {
    "@type": "Question",
    "name": "How long does it take to build a website?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "A typical showcase website takes 6 to 10 weeks from kickoff to launch. E-commerce and marketplace projects require 12 to 20 weeks."
    }
  }
]
}

Article and BlogPosting

Content markup helps AI agents assess the authority and recency of your published material. Always include author, datePublished, dateModified, and publisher properties. AI systems use these to evaluate whether your content is current and credible — a blog post from 2022 about AI visibility will be weighted far less than one from 2026.

{
"@type": "BlogPosting",
"headline": "How to implement llms.txt for your business",
"author": {
  "@type": "Person",
  "name": "Sebastien Balieu",
  "jobTitle": "Founder",
  "worksFor": { "@type": "Organization", "name": "Numinam" }
},
"datePublished": "2026-03-05",
"dateModified": "2026-03-20",
"publisher": { "@type": "Organization", "name": "Numinam" }
}

Person

Personal brand entities matter more than ever in the age of AI. AI agents evaluate content credibility by assessing the author’s identity, expertise, and linked presence across the web. A well-marked-up Person entity connected to your content establishes the E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals that both Google and AI systems value.

Include sameAs links pointing to the person’s LinkedIn profile, Twitter account, and any other authoritative profiles. This helps AI agents disambiguate the person and connect them to their broader digital footprint.

Implementation: JSON-LD format explained

JSON-LD (JavaScript Object Notation for Linked Data) is the recommended format for Schema.org markup. Google explicitly recommends it over Microdata or RDFa, and AI agents parse it most reliably. The markup goes in a <script type=“application/ld+json”> tag, typically in the <head> of your HTML document.

Key principles for AI-optimized JSON-LD:

• One comprehensive graph per page. Use @graph to combine multiple entities (Organization, WebPage, Service, BreadcrumbList) in a single JSON-LD block. This gives AI agents a complete picture in one parse.
• Use @id references to link entities. When your Service references your Organization, use “provider”: {"@id": "https://yourdomain.com/#organization"} to create a linked data graph rather than duplicating information.
• Be specific with types. Use the most specific Schema type available. ProfessionalService is better than LocalBusiness for an agency. MedicalClinic is better than LocalBusiness for a clinic.
• Include all relevant properties. AI agents build confidence scores based on data completeness. A Service with name, description, provider, areaServed, offers, and category is far more actionable than one with just a name and description.
• Keep data synchronized with visible content. Contradictions between your Schema markup and on-page content confuse both Google and AI agents — and can result in penalties or exclusion from citations.

Complete implementation example

Here is a comprehensive JSON-LD implementation for a service business home page, combining multiple entity types into a single linked graph:

<script type="application/ld+json">
{
"@context": "https://schema.org",
"@graph": [
  {
    "@type": "ProfessionalService",
    "@id": "https://example.com/#organization",
    "name": "Example Agency",
    "url": "https://example.com",
    "description": "Digital agency specializing in web development and AI visibility.",
    "address": {
      "@type": "PostalAddress",
      "addressLocality": "Brussels",
      "addressCountry": "BE"
    },
    "areaServed": ["Belgium", "France"],
    "priceRange": "EUR EUR EUR",
    "knowsLanguage": ["fr", "en"]
  },
  {
    "@type": "Service",
    "@id": "https://example.com/#service-webdev",
    "name": "Web Development",
    "provider": { "@id": "https://example.com/#organization" },
    "description": "Custom website development optimized for performance and conversion.",
    "offers": {
      "@type": "Offer",
      "priceCurrency": "EUR",
      "price": "5000"
    }
  },
  {
    "@type": "WebPage",
    "@id": "https://example.com/#webpage",
    "url": "https://example.com",
    "name": "Example Agency — Web Development & AI Visibility",
    "isPartOf": { "@id": "https://example.com/#website" },
    "about": { "@id": "https://example.com/#organization" }
  },
  {
    "@type": "BreadcrumbList",
    "itemListElement": [
      {
        "@type": "ListItem",
        "position": 1,
        "name": "Home",
        "item": "https://example.com/"
      }
    ]
  }
]
}
</script>

Testing and validation tools

Validation ensures your markup is syntactically correct and follows Schema.org specifications. Use these tools in combination:

• Google Rich Results Test (search.google.com/test/rich-results): Validates markup and shows eligible rich result types. Note this tests Google compatibility — AI agents may parse additional types Google does not support as rich results.
• Schema.org Validator (validator.schema.org): Tests against the full Schema.org specification, not just Google’s supported subset. Use this for comprehensive validation.
• Manual browser inspection: View page source and search for application/ld+json to verify the markup renders correctly in production. Check that dynamic rendering (SSR/SSG) outputs the JSON-LD as expected.
• AI agent testing: Ask ChatGPT or Perplexity about your business and compare their response against your structured data. If the AI gets facts wrong that are correctly defined in your Schema, the markup may not be deployed correctly or the page may not be indexed yet.
• Structured Data Linter (linter.structured-data.org): Catches common errors like missing required properties, incorrect value types, and deprecated terms.

Common mistakes to avoid

• Implementing Schema only for Google rich results. Many businesses only add the minimum markup needed to trigger rich snippets (Review, FAQ, Product). For AI agents, you need a complete entity profile — Organization, Service, Offer, Person, and content types working together.
• Duplicating rather than linking entities. If your Organization appears on every page, use @id references instead of repeating the full entity. Duplicated entities with slight variations confuse AI parsers and create inconsistencies.
• Missing pricing information. AI agents performing comparisons heavily weight pricing data. If competitors include PriceSpecification and you do not, agents have less confidence in recommending you — they cannot compare what they cannot quantify.
• Outdated markup. Schema data that contradicts your current offerings — old prices, discontinued services, wrong addresses — actively harms your AI visibility. Treat structured data as a living asset that requires regular maintenance.
• Ignoring author markup. For blog posts and articles, missing Person entities for authors weakens E-E-A-T signals. AI agents evaluate content credibility partly through author identity and expertise.
• Using Microdata or RDFa instead of JSON-LD. While technically valid, these inline formats are harder for AI systems to parse and more prone to implementation errors. JSON-LD is the industry standard.
• Inconsistent naming. If your Schema says “Example Agency” but your website header says “Example Digital” and your Google Business Profile says “Example Agency Inc.”, AI agents struggle to build a coherent entity. Use the exact same name everywhere.

Schema markup in your broader AI visibility strategy

Structured data does not exist in isolation. It is one layer of a comprehensive AI visibility strategy:

• llms.txt — a curated Markdown summary of your business for AI consumption, providing the narrative context that Schema alone cannot convey.
• Agentic GEO — making your website actionable by autonomous AI agents through structured APIs, clear calls-to-action, and machine-readable workflows.
• Brand entity presence — establishing your business identity across platforms that AI models use for knowledge: LinkedIn, YouTube, Wikipedia, industry directories.
• Technical performance — fast-loading pages that AI crawlers can access and parse efficiently, because a page that takes 8 seconds to load may be skipped entirely by a time-constrained AI agent.

Schema markup is the foundation because it provides the structured facts that every other layer builds upon. Without it, AI agents are left guessing — and when they guess, they often choose a competitor who made understanding easy.

Key takeaways

• Schema markup is no longer optional for AI visibility. AI agents rely on structured data to understand, compare, and recommend businesses — not just to display rich snippets.
• Focus on the core types: Organization/ProfessionalService, Service, Offer, PriceSpecification, FAQPage, BlogPosting, Article, and Person.
• Use JSON-LD format with @graph and @id references for clean, linked data that AI agents can parse in a single pass.
• Be comprehensive. More complete data builds higher confidence scores with AI agents — include pricing, service areas, languages, and author information.
• AI agents use structured data for decisions, not display. Incomplete or inaccurate markup carries higher consequences than missing a rich snippet.
• Validate with Google Rich Results Test, Schema.org Validator, and Structured Data Linter, then test with actual AI agents to verify real-world impact.
• Treat structured data as a living asset — review quarterly, synchronized with your website content and business changes.