/ BLOG17 May 202610 min

Schema markup that Google AI Overviews cites: real composition that works

How to combine Article, FAQPage, HowTo and BreadcrumbList to maximize citations in AI Overviews and ChatGPT. Real examples, not schema.org copy-paste.

schema-markup
structured-data
ai-overviews
technical-seo

"Add Schema" is the most repeated and worst-explained SEO advice of recent years. Most sites implement a flat Article or a generic LocalBusiness, then wonder why Google AI Overviews and ChatGPT don't cite them.

This post is what actually works in 2026, with real examples of Schema composition I see cited in AI responses, and the anti-patterns I keep finding in sites that come to audit.

The important change: Schema is no longer just for rich snippets

Until 2023, Schema mainly served to get rich snippets (stars, FAQ dropdowns, visual breadcrumbs) in SERPs. In 2026 there's a second function that's equally or more important:

Generative AIs use Schema to understand what each page is and decide what to cite.

ChatGPT, Perplexity and Google AI Overviews read JSON-LD before the visible content. If your Schema is vague or incorrect, they ignore you. If your Schema is specific and rich, they pick you among 5 alternatives with equivalent content.

Schema composition by page type

Informative blog post

Wrong (what I see in 70% of WP blogs):

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "My post",
  "datePublished": "2026-05-01"
}

Right (rich composition):

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Article",
      "@id": "https://example.com/blog/my-post/#article",
      "headline": "How to appear in ChatGPT in 2026",
      "description": "Practical guide with real data on Bing indexing, Schema and topical authority.",
      "datePublished": "2026-05-17T10:00:00+02:00",
      "dateModified": "2026-05-17T10:00:00+02:00",
      "wordCount": 2100,
      "inLanguage": "en-GB",
      "image": "https://example.com/blog/my-post/hero.webp",
      "author": { "@id": "https://example.com/#person" },
      "publisher": { "@id": "https://example.com/#person" },
      "mainEntityOfPage": "https://example.com/blog/my-post/",
      "isPartOf": { "@id": "https://example.com/blog/#website" }
    },
    {
      "@type": "Person",
      "@id": "https://example.com/#person",
      "name": "Jesús Porres",
      "url": "https://example.com/about",
      "sameAs": [
        "https://www.linkedin.com/in/jesusporres",
        "https://github.com/jesusporres"
      ],
      "jobTitle": "SEO Consultant",
      "knowsAbout": ["Technical SEO", "AI applied to SEO", "Schema markup"]
    },
    {
      "@type": "BreadcrumbList",
      "itemListElement": [
        { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://example.com/" },
        { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://example.com/blog/" },
        { "@type": "ListItem", "position": 3, "name": "How to appear in ChatGPT" }
      ]
    }
  ]
}

What this composition adds:

@graph groups multiple related entities in a single structure.
Unique @id lets nodes reference each other (Article points to Person via @id).
Person with sameAs triangulates authority: AIs cross-check your LinkedIn, GitHub, etc.
BreadcrumbList helps hierarchical understanding of the site.
knowsAbout on Person labels you as a topical expert.

Post with FAQs

Add FAQPage when the post explicitly answers questions. Do NOT invent FAQs just to add the Schema — AIs detect over-optimization.

{
  "@type": "FAQPage",
  "@id": "https://example.com/blog/my-post/#faq",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does it take for a site to appear in ChatGPT?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Between 2 and 6 weeks after correctly indexing in Bing Webmaster Tools and activating IndexNow. New sites can take up to 8 weeks."
      }
    },
    {
      "@type": "Question",
      "name": "Do you need to pay to appear in ChatGPT?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "No. ChatGPT has no system for paid mentions. Anyone selling «guaranteed citations» is selling smoke."
      }
    }
  ]
}

Bonus: each Answer.text ideally between 80-200 characters. AIs prefer direct and verbatim-citable answers.

Step-by-step tutorial

HowTo still works even though Google removed the visual rich snippet in 2023. AIs still read it.

{
  "@type": "HowTo",
  "name": "How to register your site in Bing Webmaster Tools",
  "totalTime": "PT15M",
  "estimatedCost": { "@type": "MonetaryAmount", "currency": "EUR", "value": "0" },
  "step": [
    {
      "@type": "HowToStep",
      "position": 1,
      "name": "Create account",
      "text": "Go to webmaster.bing.com and create an account with your email.",
      "url": "https://webmaster.bing.com/"
    },
    {
      "@type": "HowToStep",
      "position": 2,
      "name": "Add your site",
      "text": "Click «Add a site», enter your site URL, verify ownership via meta tag or DNS TXT."
    },
    {
      "@type": "HowToStep",
      "position": 3,
      "name": "Submit sitemap",
      "text": "In Sitemaps, paste your sitemap.xml URL (typically https://yourdomain.com/sitemap.xml)."
    }
  ]
}

Service page

For a service page (commercial landing):

{
  "@context": "https://schema.org",
  "@type": "Service",
  "@id": "https://example.com/services/local-seo/#service",
  "name": "Local SEO for small businesses",
  "description": "Local SEO service for small businesses in Spain with initial setup + monthly retainer.",
  "provider": {
    "@type": "LegalService",
    "@id": "https://example.com/#legalservice",
    "name": "Jesús Porres SEO Consulting",
    "url": "https://example.com/",
    "telephone": "+34 ...",
    "address": { ... },
    "priceRange": "€€"
  },
  "areaServed": {
    "@type": "Country",
    "name": "Spain"
  },
  "serviceType": "Local SEO",
  "offers": {
    "@type": "AggregateOffer",
    "priceCurrency": "EUR",
    "priceSpecification": {
      "@type": "PriceSpecification",
      "priceCurrency": "EUR",
      "minPrice": "800"
    }
  },
  "hasOfferCatalog": {
    "@type": "OfferCatalog",
    "name": "Included services",
    "itemListElement": [
      { "@type": "Offer", "itemOffered": { "@type": "Service", "name": "Initial GBP audit" }},
      { "@type": "Offer", "itemOffered": { "@type": "Service", "name": "LocalBusiness Schema" }},
      { "@type": "Offer", "itemOffered": { "@type": "Service", "name": "Monthly GBP posts" }},
      { "@type": "Offer", "itemOffered": { "@type": "Service", "name": "Monthly report" }}
    ]
  }
}

hasOfferCatalog moves the needle most. Almost no one implements it well.

The 5 anti-patterns I keep seeing

1. Schema only on home

I see this on 60% of sites. Home with complete Schema, rest of pages empty. Each page must have its specific Schema for its type.

2. JSON-LD with syntax errors

Trailing commas, mismatched quotes, keys without quotes. Any syntax error makes the browser parse JSON-LD as invalid and Google ignores it all. Validate with Google's Schema Validator (the official one), not Rich Results Test.

3. `@type` that doesn't exist in schema.org

Inventing types like "@type": "ConsultingService" (doesn't exist). The correct one is "@type": "ProfessionalService" with serviceType: "Consulting". If in doubt, schema.org has a search box.

4. Contradictory data between Schema and visible page

If Schema says priceRange: "€€" and the page shows services at 3,000€/month, there's contradiction. AIs penalize contradiction.

5. Multiple Schemas of the same type on one page

Marking the same thing twice (e.g. two different Organization on home). Google keeps one arbitrarily. Better one @graph with all unique entities.

How to actually validate

Official Schema Validator (https://validator.schema.org/). The only one that validates against the full spec.
Rich Results Test only validates types with visual rich snippet (limited).
Google Search Console → Enhancements → Structured data shows errors in real coverage.
For mass audits, Screaming Frog has a structured data validation module in licensed plan.

Recommended workflow when launching a new page

Decide the main @type based on the page type (Article, Service, Product, etc.).
Build the JSON-LD locally.
Validate with Schema Validator before deploy.
After deploy, verify the JSON-LD renders correctly with View Source.
Submit the URL in Google Search Console → URL Inspection to force re-crawl.
Wait 7-14 days and check in GSC → Enhancements if the type is recognized.

Why it matters more in 2026

Three cumulative reasons:

AI Overviews in global rollout, including Spain progressively. Rich Schema = higher chance of citation.
ChatGPT with search enabled by default. Citations in responses are the new SERP impression.
Perplexity and other LLMs value structured extraction over pure prose.

Schema markup done right is the best-ROI technical investment I know of in 2026. Cost: 3-5 hours per main page of the site. Benefit: months of AI citations + rich snippets + better understanding by Google.

Want to review your site's current Schema with external criteria? It's a standard part of the technical SEO audit. For implementation + monthly monitoring, see technical SEO.

[ SEGUIR LEYENDO ]