Structured Data for AI Search: The Schema Types That Actually Get Cited

Structured data for AI search means marking up a page with Schema.org JSON-LD so ChatGPT, Perplexity, and Google's AI Overviews can identify what the content actually is: an article, an organization, a product, a set of steps. A handful of types do most of the work (Organization, Article, Person, BreadcrumbList), while others like FAQPage and HowTo only matter when the page genuinely fits them. Schema helps engines understand and trust a page; it does not force a citation on its own.

Key takeaways

• Structured data does not directly cause AI citations, but it removes ambiguity about what a page is, which is a precondition for being trusted enough to cite.
• Ahrefs found that 53% of pages cited by AI Overviews carry schema markup, about three times the rate on pages that are not cited, though the study could not prove causation.
• Organization, Article/BlogPosting, Person, and BreadcrumbList are the schema types worth prioritizing on almost every page.
• FAQPage and HowTo are situational: FAQPage no longer earns a Google rich result as of May 2026, and HowTo only fits genuine step-by-step content.
• Google's crawlers read JSON-LD at index time, but Search Engine Land has reported that some chatbots fetch pages live during a conversation and read the visible text instead, so clean writing still matters more than markup in that moment.
• Match the schema type to what the page truly is. Mismatched or padded markup is a trust problem, not a shortcut.

What is structured data, and why does AI search use it?

Structured data is a shared vocabulary, defined at Schema.org, that labels the parts of a page in a format machines can parse without guessing. Instead of a crawler inferring from paragraph text that a page is a recipe, a local business, or a how-to guide, JSON-LD states it directly: this is an Article, published on this date, written by this Person, published by this Organization. On a typical page the JSON-LD sits in a script tag in the head, separate from the text a visitor reads.

AI search systems inherited this need from traditional search. Google's crawlers already parse JSON-LD to power rich results in classic Search, and that same index feeds Google's AI Overviews and AI Mode. Standalone AI products such as ChatGPT and Perplexity build their own understanding of the web partly through crawling and partly through live retrieval when a user asks a question, and structured data plays a different role at each stage, which is the distinction most guides on this topic skip.

How this differs from the WordPress schema setup question

If you have not added schema markup to your site yet, start with the mechanics: our guide on how to add schema markup in WordPress covers the plugin and manual JSON-LD routes. This page assumes the markup exists and answers a different question: which types are actually worth your time for AI search, and which ones are decoration.

Which schema types actually get cited in AI search?

Organization, Article or BlogPosting, Person, and BreadcrumbList carry the most weight, because together they answer the questions every AI system needs settled before it treats a page as a citable source: who published this, what type of content is it, who wrote it, and where does it sit in the site. Type-specific schema such as Product, Review, Event, or HowTo only adds value when the page is genuinely that kind of content; forcing it onto a page that is not creates a mismatch rather than an advantage.

This is not a guess. Google Search Central's own documentation recommends JSON-LD generally and treats Organization, Article, and author information as foundational identity signals, and Search Engine Land's 2026 analysis of schema for AI search reaches the same conclusion: entity definition and entity relationships (who made this, who is it about, how does it connect to the rest of the site) matter more for AI understanding than niche, page-specific types.

Does adding schema markup increase AI citations?

Not reliably, according to the largest test done on this so far. Ahrefs tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026 and compared them against roughly 4,000 control pages already receiving AI citations. The result: Google AI Overview citations changed by -4.6% (a small but statistically significant decline), AI Mode changed by +2.4%, and ChatGPT changed by +2.2%, with the latter two statistically indistinguishable from zero.

The same study found that 53% of AI-cited pages already carried schema, about three times the rate among pages that were not cited. Ahrefs was careful about the interpretation: because every page in the test already had 100-plus AI Overview citations before any schema was added, the correlation likely reflects that well-built, well-established sites both rank well and tend to have schema, not that schema itself pulled in the citations. Treat structured data as infrastructure that supports a page's credibility, not a lever that manufactures it.

Does FAQPage schema still matter after Google removed FAQ rich results?

FAQPage schema can still help search and AI systems parse question-and-answer content, but it stopped earning the expandable FAQ display in Google Search. Google announced on May 7, 2026 that FAQ rich results would no longer appear, and by June 15, 2026 it had removed the FAQPage documentation from Search Central's rich-result guidance entirely. Sites do not need to remove existing FAQPage markup, but adding it purely to chase the old visual snippet no longer makes sense.

The practical takeaway: keep FAQPage markup where you already publish genuine FAQs, since it mirrors how AI systems like clean, extractable question-and-answer pairs, including the FAQ block later on this page. Just do not treat it as a growth tactic on its own, and never pad a page with invented questions solely to make the block bigger.

How do AI engines like ChatGPT, Perplexity, and Google AI Overviews actually use structured data?

Google's AI Overviews and AI Mode draw on the same index that Google Search has always built from crawled pages, so JSON-LD parsed at crawl time feeds directly into how Google's systems understand and rank a page for those AI features. Standalone chatbots behave differently in the moment a user asks a question: Search Engine Land's 2026 reporting notes that several AI systems fetch a page live during a conversation and extract the visible HTML text rather than reading the JSON-LD block at that instant, meaning clear writing and accurate on-page facts matter as much as, or more than, the markup during that live fetch.

The two behaviors are not in conflict. Structured data still shapes how a page gets indexed and understood over time, which is the foundation every AI system eventually draws on, even if a live fetch skips straight to the body text. Building both, clean markup and clean, well-organized prose, covers each stage instead of betting on one.

What structured data actually does for AI search

Three jobs, not one. Structured data states entities plainly so an engine does not have to infer them from prose. It separates that plain statement from the visible text, so a parser reading the head does not have to untangle marketing copy to find a publish date. And across the datasets researchers have published, it correlates with the kind of established, well-maintained sites that AI systems already trust, even if it is not the cause of that trust. Understanding which of the three jobs you are solving for tells you whether a given schema type is worth adding to a specific page.

How to add the schema types that matter for AI search

You do not need every type on every page. Work through this order and stop once a type does not apply.

1. Add Organization schema sitewide. Legal name, logo, and sameAs links to your verified social and business profiles. This runs once, in a template, and covers the whole site.
2. Add Article or BlogPosting to every post. Headline, author, datePublished, dateModified, and publisher. If you run WordPress, an SEO plugin like Rank Math or Yoast generates this automatically; see our WordPress schema markup guide for the setup steps.
3. Add Person schema to bylines. Connect the author name in your Article schema to a Person object with a real jobTitle and, where possible, a sameAs link to a verifiable profile.
4. Add BreadcrumbList to every page below the homepage. Low effort, and it gives both Google and AI crawlers an explicit map of where the page sits.
5. Add LocalBusiness only if you have a physical location. Name, address, phone, and hours, matched exactly to what is on the page and in your Google Business Profile.
6. Add Product, Review, Event, or HowTo only when the page truly is that thing. Product schema on a blog post, or HowTo on a listicle with no real steps, is the most common way sites turn structured data into a liability instead of an asset.
7. Validate before you trust it. Run the page through Google's Rich Results Test and the Schema Markup Validator at validator.schema.org, then watch the structured data report in Search Console after the next crawl.

None of this replaces the actual content work. Structured data describes what is already on the page; it cannot substitute for depth, accuracy, or a genuinely useful answer. For the broader discipline this sits inside, see our guides on answer engine optimization and a full AEO strategy.

Common mistakes when adding schema for AI search

• Mismatched types. Marking a blog post as Product, or a service page as an Article, tells engines something false about the page. Match the type to what the page genuinely is.
• Padding FAQPage with invented questions. Writing fake questions just to fill out a schema block is both a poor reader experience and a markup that no longer pays off in Google's rich results anyway.
• Skipping validation. Invalid JSON-LD is often silently ignored rather than flagged, so a broken block can sit unnoticed for months. Test every new template.
• Running two schema sources at once. Two SEO plugins, or a plugin plus manual JSON-LD, both trying to output the same type creates duplicate or conflicting markup.
• Treating schema as a citation shortcut. The Ahrefs data above is clear: schema alone did not move citations in a controlled test. Pair it with genuinely complete, well-sourced content.
• Letting the data go stale. If a datePublished, price, or author field no longer matches the live page, fix or remove it. Wrong structured data is worse than none.

Frequently asked questions

What is structured data for AI search? Structured data for AI search is Schema.org markup, usually written as JSON-LD, that tells search engines and AI systems what a page actually is: an article with an author, an organization with a real address, a product with a price. AI Overviews, AI Mode, and chatbots like ChatGPT and Perplexity use it as one input among many when deciding what to cite.

Which schema type matters most for AI search visibility? Organization and Article or BlogPosting matter most, because they establish who published the content and what type of content it is, which every AI system needs before it will trust and cite a source. Person schema on the author and BreadcrumbList for site structure come next. Type-specific schema like Product or Event only matters when the page genuinely is that thing.

Does FAQPage schema still help with AI search? It can still help AI systems parse question-and-answer content, but it no longer earns the expandable FAQ rich result in Google Search. Google announced on May 7, 2026 that FAQ rich results would stop appearing, and removed the documentation in June 2026. Keep FAQPage schema for clarity, not for a rich-result payoff.

Do I need schema markup to be cited by ChatGPT or Perplexity? No. Search Engine Land has reported that several AI systems extract visible HTML text during live retrieval and do not read JSON-LD at that moment, so clear, well-written content matters more than markup for a same-session citation. Schema still helps in the earlier crawl-and-index stage that feeds a system's broader knowledge.

What is the difference between schema for Google AI Overviews and schema for chatbots? Google's crawlers read JSON-LD at crawl time and feed it into the same index that powers AI Overviews and AI Mode, so standard Schema.org markup applies directly. Standalone chatbots such as ChatGPT and Perplexity often fetch pages live during a conversation and lean on the visible page text, so the markup matters less at that exact moment but still supports how the page was indexed originally.

Does adding schema markup guarantee an AI citation? No. Ahrefs tracked 1,885 pages that added JSON-LD schema and found no meaningful citation increase: Google AI Overview citations changed by -4.6%, AI Mode by +2.4%, and ChatGPT by +2.2%, all statistically indistinguishable from zero except the AI Overview decline. Schema supports clarity and trust; it does not force a citation on its own.

Should I add HowTo schema for AI search? Only if the page walks through genuine ordered steps, such as a real tutorial. Google's own guidance treats HowTo as a page-specific type, not a general AI-visibility booster, and forcing it onto content that is not actually a procedure creates markup that misrepresents the page, which can hurt more than help.

What is the best format for structured data: JSON-LD, Microdata, or RDFa? JSON-LD. Google Search Central recommends it explicitly, describing it as the easiest format for site owners to implement and maintain at scale and less prone to errors than Microdata or RDFa, since the data sits in a separate script block instead of being woven into the visible HTML.

How do I test whether my structured data is valid? Run the live URL through Google's Rich Results Test to check eligibility for Google's supported types, then check the broader Schema.org vocabulary with the Schema Markup Validator at validator.schema.org. After publishing, watch the structured data reports in Search Console for errors that only appear once Google recrawls the page.

Can too much schema markup hurt my page? Markup that misdescribes a page, such as Product schema on a blog post or a FAQPage block padded with questions nobody asks, creates a mismatch between what the schema claims and what the page delivers. That mismatch is a trust problem for both Google and AI systems. Mark up only what the page genuinely is, and keep the data accurate.

What to do next

Start with Organization and Article schema sitewide, add Person and BreadcrumbList, and only reach for Product, HowTo, or FAQPage where the page genuinely fits. Validate every template, then move your attention to the content itself, since structured data can only clarify what is already strong. If you want a full audit of your structured data and where your AI-search visibility is actually leaking, request a free SEO audit from Rankite.