Schema markup for AI search: a 2026 field guide

The schema types that move the needle for AI search citation in 2026 are Organization, Service with Offer, FAQPage, Article, and Person. Most plugins emit bloated WebPage and BreadcrumbList schema that LLMs ignore. The ones that matter need to be added manually or through a lightweight theme integration. This is the working field guide.

If your SEO checklist still says “install Yoast and you’re done,” you are missing roughly 70 percent of the schema signal that AI search products actually weight. Yoast Free emits WebSite, BreadcrumbList, and a basic WebPage object on every page. That is not enough.

What schema does for AI citation (and what it doesn’t)

Schema markup is JSON-LD structured data that explicitly describes what’s on a page. LLMs use schema as one of several signals to determine entity coherence, content category, and citation eligibility.

The clearest signal in 2025-2026 data: pages with valid schema for the content type get cited at meaningfully higher rates than pages without. Profound’s open citation index showed roughly a 1.8x lift on FAQPage schema versus pages without it for query categories where the FAQ format was relevant.

What schema does NOT do: it does not magically rank a low-quality page. The content underneath still needs to answer the question. Schema makes good content more findable by AI. It does not make bad content credible.

The five schema types that matter most

1. Organization schema (every page, in the site-wide header)

Organization is the most important schema for entity recognition. LLMs use it to determine “is this a real, coherent business that I should trust as a source.”

Minimum required fields for AI citation:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://yourdomain.com/#organization",
  "name": "Your Business Name",
  "url": "https://yourdomain.com/",
  "email": "hello@yourdomain.com",
  "description": "One sentence that names what you do.",
  "sameAs": [
    "https://www.linkedin.com/company/yourbusiness",
    "https://twitter.com/yourbusiness",
    "https://github.com/yourbusiness"
  ],
  "parentOrganization": {
    "@type": "Organization",
    "name": "Parent Company Name"
  }
}

The sameAs array is the underrated signal. It tells the LLM “these external profiles are also this entity.” When all your sameAs URLs describe the business consistently, the entity coherence score goes up. When they conflict, the score drops.

2. Service schema with Offer (on service pages)

If you sell a service with a price, Service schema with embedded Offer schema is a huge AI search signal. Generative engines that get asked “what does [your service] cost” can extract directly from this.

{
  "@type": "Service",
  "@id": "https://yourdomain.com/#service",
  "name": "Your Service Name",
  "serviceType": "Specific Service Category",
  "provider": { "@id": "https://yourdomain.com/#organization" },
  "areaServed": "Worldwide",
  "description": "What the service does and who it's for.",
  "offers": [
    {
      "@type": "Offer",
      "name": "Tier 1 Name",
      "description": "What's included in Tier 1.",
      "price": "497",
      "priceCurrency": "USD",
      "priceSpecification": {
        "@type": "UnitPriceSpecification",
        "price": "497",
        "priceCurrency": "USD",
        "unitText": "MONTH"
      }
    }
  ]
}

The priceSpecification with unitText: "MONTH" is the part most schema implementations miss. It tells the engine this is a monthly subscription, not a one-time price. The accuracy of AI answers about your pricing goes up dramatically when this is correct.

3. FAQPage schema (on every page with a real FAQ)

FAQPage is the highest-leverage single schema for AI citation. It works for two reasons. First, it directly maps to how generative engines extract Q&A pairs. Second, it makes the page eligible for FAQ rich results in classic Google search.

{
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Question your buyer actually asks",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Direct, sourced, no-qualifier answer in 60 to 120 words."
      }
    }
  ]
}

Critical implementation detail: the answer text in the schema must match (or closely match) visible HTML on the page. Google deindexed FAQ rich results for sites that put answers only in schema. LLMs that detect schema-HTML drift also discount the page.

4. Article schema (on every blog post)

Article schema tells AI engines “this is a discrete piece of content with an author, publication date, and topic.” Blog posts with valid Article schema get cited more reliably as sources for the questions they answer.

{
  "@type": "Article",
  "@id": "https://yourdomain.com/your-post/#article",
  "headline": "Your post headline",
  "description": "One-sentence summary of the post.",
  "datePublished": "2026-06-02",
  "dateModified": "2026-06-02",
  "author": { "@id": "https://yourdomain.com/#organization" },
  "publisher": { "@id": "https://yourdomain.com/#organization" },
  "mainEntityOfPage": "https://yourdomain.com/your-post/"
}

Update dateModified whenever you revise the post substantively. This is one of the freshness signals AI engines use.

5. Person schema (on your About page)

Person schema for the founder, lead consultant, or face of the business creates an entity link between your Organization and a named human. This matters for trust signals in regulated or expert-led categories (legal, medical, finance, B2B services).

{
  "@type": "Person",
  "@id": "https://yourdomain.com/#founder",
  "name": "Full Name",
  "jobTitle": "Job Title",
  "worksFor": { "@id": "https://yourdomain.com/#organization" },
  "sameAs": [
    "https://www.linkedin.com/in/yourprofile",
    "https://twitter.com/yourhandle"
  ]
}

The schema types most plugins emit (and why they’re not enough)

Yoast SEO Free emits WebSite, BreadcrumbList, and WebPage schema on every page. These are useful for classic Google search but provide minimal lift for AI citation.

• WebSite — establishes the site as a discrete entity. Marginal lift for citation.
• BreadcrumbList — helps Google understand site hierarchy. Mostly irrelevant for AI extraction.
• WebPage — generic page schema. LLMs largely ignore it in favor of more specific types.

The schemas that drive AI citation (Organization, Service+Offer, FAQPage, Article, Person) require either manual implementation in your theme or a more capable plugin (RankMath Free can do FAQPage; Yoast Premium adds Person, WebSite Author, and others).

Our practical recommendation: implement Organization + Service + FAQPage in your theme’s wp_head hook directly. It’s about 200 lines of PHP, runs faster than a plugin, and gives you full control over the JSON-LD output.

How to test if your schema is working

Three free tools:

• Google Rich Results Test — paste any URL and see which Google-eligible schemas it detects. Validates structure.
• Schema.org Validator — paste your JSON-LD and check for syntactic errors and warnings.
• Google Search Console “Enhancements” report — shows you which schema types Google has indexed across your site and flags errors.

For AI citation specifically, the tools are newer and rougher: Profound’s citation explorer, Otterly’s tracker, and manual sampling. Run your top 10 queries through ChatGPT, Perplexity, and Gemini once a month and record which of your pages appear in source lists. That gives you a baseline.

Common schema mistakes that hurt AI citation

• Schema in the page body, not the head. Some engines only parse schema in <head>. Put it there.
• Multiple conflicting schema blocks. If three plugins each emit their own Organization schema, the engines pick one (sometimes the wrong one). Audit and consolidate.
• FAQPage schema with answers that don’t match the visible page. Both Google and the LLMs detect this and discount the page.
• Service schema without Offer. If you sell a service with a price, the Offer is the part that gets quoted. Include it.
• Stale dates. An Article with dateModified from 2022 looks abandoned. Update it whenever you refresh.
• Missing sameAs links. Without them, the LLM has to guess whether your LinkedIn and your website are the same entity. Give it explicit signal.

The honest take on schema in 2026

Schema is necessary but not sufficient. Pages with great schema and bad content still don’t get cited. Pages with great content and no schema do get cited eventually, but slower.

The fastest lift comes from pairing both. Real, sourced, well-written content with the five schemas listed above gets cited 3 to 5 times faster than the same content with default plugin schema. That’s the actual leverage.

If you’re going to spend an hour on schema this quarter, spend it adding FAQPage with three real buyer questions to your top five pages, plus Service+Offer schema on your pricing page if you have one. That alone moves the citation needle on most sites.