How to Use Schema Markup for AI Search in 2026: What to Implement and What the Evidence Says

Schema markup is worth getting right, but it is not a demonstrated way to get cited. Ahrefs found no clear citation uplift when already-cited pages added JSON-LD, in its matched analysis. Google says there is no special schema.org structured data to add to appear in its AI features. And in searchVIU's October 2025 test of five AI systems, none returned a price that existed only in JSON-LD. This guide shows which markup to implement, how to implement it correctly, and what the evidence does not support.

What the Evidence Says About Schema and AI Search

Ahrefs' matched study followed pages that added JSON-LD between August 2025 and March 2026 against matched control pages (1,885 treated pages and 4,000 control pages). Relative to controls, citations moved +2.4% on Google AI Mode and +2.2% on ChatGPT, both indistinguishable from zero, and fell 4.6% on Google AI Overviews, a small but statistically significant decline that Ahrefs says it cannot definitively pin on schema. Every page already had at least 100 AI Overview citations in February 2025, so the study speaks to adding schema to pages that are already being cited. Ahrefs says schema might still help pages AI systems are not seeing at all, and that its data cannot show it.

searchVIU's tests put prices and product details in visible HTML, JavaScript-rendered content and several structured-data formats. None of the five systems returned a price that appeared only in JSON-LD. ChatGPT and Gemini returned prices that were in visible HTML, Claude returned none of the eight prices, and Perplexity and Google AI Mode answered mainly from their indexes. searchVIU separates this direct-fetch step from training, indexing and search, and says markup could be used in those earlier phases, especially by systems with access to a search index.

We know of no study showing that schema completeness raises AI citation. In Geology's B2B SaaS study, schema markup correlated with AI citation at r=0.103, close to no relationship in that sample. In Ahrefs' May 2025 AI Overviews study of 75,000 brands, branded web mentions correlated with AI Overview brand visibility at 0.664 against 0.218 for backlinks; in its December 2025 follow-up across ChatGPT, AI Mode and AI Overviews, YouTube mentions correlated at about 0.737. Those are correlations with brand-name mentions anywhere on the web, not a measure of earned media, and correlation is not proof of cause. Treat schema as a legibility fix rather than a proven visibility lever.

Google also says correct implementation does not guarantee a rich result. The rest of this guide treats schema as accurate plumbing.

The Schema Types Worth Implementing

Schema.org defines hundreds of types. Use the ones that describe what is actually on your pages. We found no study that ranks types by citation effect, so the descriptions below say what each type declares and when to use it.

1. Organization Schema

Organization markup on your homepage states who the site belongs to, its logo and contact details, and which other profiles describe the same entity.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Northwind",
  "url": "https://example.com",
  "logo": "https://example.com/logo.png",
  "description": "A company whose public description states what it does, who it serves, and where to verify the claim.",
  "sameAs": [
    "https://www.linkedin.com/company/example"
  ]
}

Use the same name, description and profile links the site and profiles already show. If the facts differ between your site and your profiles, fix the facts first.

2. Article / BlogPosting Schema

Article markup declares the headline, author, dates and image for each post. Google's Article documentation lists no required properties and asks you to add the ones that apply.

{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "Your Title Here",
  "author": {
    "@type": "Person",
    "name": "Author Name",
    "url": "https://example.com/about/author"
  },
  "publisher": {
    "@type": "Organization",
    "name": "Northwind"
  },
  "datePublished": "2026-01-15",
  "dateModified": "2026-01-15",
  "image": "https://example.com/images/post.png"
}

articleBody is an optional property. If you include it, it must match the visible text. Do not truncate it to fit a theory about what AI needs: the tests above found chatbots reading visible text, and we found no study showing that a longer articleBody changes citation.

3. Product Schema (only if you sell a product)

If you sell a product or service with a public price, Product markup states the offer. Only mark up prices, ratings and reviews that are real and visible on the page.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Example Product",
  "description": "What the product is, in one sentence.",
  "brand": {
    "@type": "Brand",
    "name": "Northwind"
  },
  "offers": {
    "@type": "Offer",
    "priceCurrency": "USD",
    "price": "99.00",
    "availability": "https://schema.org/InStock"
  }
}

The values above are placeholders, not anyone's price or rating. Do not add ratings or review counts you cannot show on the page. Google's guidelines say not to mark up content readers cannot see.

4. FAQPage Schema

FAQPage markup should mirror a question-and-answer section readers can see. Google stopped showing FAQ rich results on May 7, 2026, so any FAQ markup you keep has to match the visible page. Whether an answer is useful depends on the answer: a citation absorption study found Q&A formatting alone does not improve how much a cited page contributes to an answer.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Your visible question?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "The same answer text a reader sees on the page."
      }
    }
  ]
}

How to Implement Schema Correctly

Step 1: Choose JSON-LD

Google recommends JSON-LD where your setup allows it. Place it in a <script type="application/ld+json"> tag in the <head> or <body>.

Step 2: Add the properties that apply

Include the properties that describe your content, and nothing you cannot show on the page.

Step 3: Validate

Use Schema.org's validator and Google's Rich Results Test. Fix syntax errors, invalid dates, broken image or logo URLs and malformed JSON. We found no evidence about how AI systems treat markup with errors, but errors make the markup unreliable for any consumer.

Step 4: Keep markup and the page aligned

Do not mark up content that is not visible. Update the markup when the page changes, including dateModified when the content has actually changed.

Schema and Authority Are Separate Problems

Schema describes your page. Coverage by independent, credible publications is a different input, and it is the one a brand cannot create on its own site. We expect third-party coverage to matter more than markup for whether an answer names a brand, but that is a reasoned expectation: the Ahrefs correlations above concern brand mentions generally, not earned media, and no study we found measures the effect of any single placement. Do not read a placement as a guaranteed citation.

A workable order of operations: make sure the pages that matter are indexable and answer the buying question with visible evidence, keep accurate markup on them, and then invest in earned coverage that gives other sources a reason to describe your company. Track whether the brand appears in the answers you care about before and after, and credit the change to what you can show.

Common Schema Mistakes

Mistake 1: Marking up content that is not visible

If your FAQ markup lists ten questions and only three are visible on the page, the markup does not match the page. Schema should describe what readers can see.

Mistake 2: Inconsistent entity names

If the homepage Organization markup says one name, blog posts another and Product markup a third, people and crawlers have to guess whether they are the same company. Pick one canonical name and use it everywhere.

Mistake 3: Selling schema as the result

Be wary of any proposal that promises more AI citations for adding JSON-LD. The best matched evidence we found shows no clear uplift for already-cited pages. Fund a documented access, answer or authority gap instead, and measure the result.

Mistake 4: Stale or invented values

Do not leave placeholder prices, ratings, dates or author details in live markup. Update dateModified only when the content changes.

How to Measure the Effect

Schema is hard to measure in isolation, so measure the brand's presence in answers and compare before and after.

Method 1: Direct AI Search Testing

Ask ChatGPT, Perplexity and Gemini the questions your content answers, on a schedule, and record whether your brand or page appears.

Method 2: AI Citation Monitoring

An AI visibility audit or tracking tool shows when and where your brand appears in AI answers. Track that over time as you change markup, pages and coverage, and note that several changes at once cannot be separated.

Method 3: Referral Traffic from AI

Check your analytics for referrals from chatgpt.com, perplexity.ai and gemini.google.com. Rising referrals are consistent with progress but do not isolate what caused it.

Frequently Asked Questions

What is schema markup and does it matter for AI search?

Schema markup is structured data, usually JSON-LD, that states what a page and its entities are. No study we found shows that adding it raises AI citations. Ahrefs' matched study of already-cited pages found no clear uplift, and none of the five systems in searchVIU's test returned a price held only in JSON-LD. Accurate markup is reasonable upkeep.

Which schema types are most important for AI search visibility?

We found no study that ranks schema types by effect on AI citation. This guide covers Organization, Article or BlogPosting, Product when you sell a product with a public price, and FAQPage. If you add any of them, keep it accurate and matched to the visible page.

Is schema markup enough to be cited in AI search?

No. No study we found shows a citation lift from adding it. Google says a page must be indexed and eligible to appear in Search with a snippet to be a supporting link in its AI features, and markup does not change that. Coverage by credible third parties is a separate input that markup cannot supply.

How do I implement schema markup?

Use JSON-LD, include only properties that describe your visible content, validate with Google's Rich Results Test and the Schema.org Validator, and keep the markup in step with the page.

Where to Start

1. Make the page work without markup: indexable, with the answer and the evidence visible.
2. Add accurate Organization and Article markup, and validate it.
3. Add Product markup only where the page shows the same content. If the page already has a visible Q&A section and you mark it up, the markup must match it.
4. Spend the real budget on the gap: a weak answer, an access problem or missing third-party coverage.
5. Track presence in the answers you care about, and credit changes carefully.

For the budget decision behind this guide, read what Ahrefs' schema study changes. Ready to see where your brand appears in AI answers? Run a free AI visibility audit.