Structured data is the fastest path to getting cited by AI search engines. Not the only path. But the fastest one, because it does something your page text cannot do on its own: it removes ambiguity.
When ChatGPT, Perplexity, or Google AI Overviews reads your page, it has to figure out what your business is, what questions you answer, and which facts it can trust. Structured data tells it all three in a format it can parse with near-perfect accuracy. No guessing. No interpretation. No filling in gaps with information from other sources that might be wrong.
This guide covers which schema types AI search engines actually read, how to implement them, and the mistakes that make them useless. Everything here is current as of August 2026.
Why structured data matters more for AI search than for Google
Traditional Google search reads your structured data to decide how to display your result. A star rating. A FAQ accordion. A recipe card. The structured data enhanced the visual, but your ranking came from other signals: backlinks, content relevance, domain authority.
AI search engines use structured data differently. They use it for extraction.
When an AI engine generates an answer, it retrieves pages, reads them, and pulls facts to synthesize a response. If your facts are buried in prose paragraphs, the model has to interpret what you meant. If your facts are in structured data, the model reads them directly. The extraction is faster, more reliable, and more likely to result in a citation.
Think about it this way. Traditional SEO structured data was about looking good in the results page. AI search structured data is about being the source the model trusts enough to cite.
The four schema types that matter for AI search
There are hundreds of schema types defined by schema.org. Most of them do not matter for AI search. These four do.
FAQPage schema
FAQPage schema wraps question-and-answer pairs in a machine-readable format. Each Question has a name (the question) and an acceptedAnswer with the text of the answer. When an AI engine retrieves your page and finds FAQPage schema, it can extract those pairs directly and cite them as answers.
This is the single highest-impact schema type for AI search. Here is why.
People use AI search engines by asking questions. Does this restaurant take reservations? What time do they open Sunday? Do they offer gluten-free options? Each of those maps directly to a FAQPage Question object. If your page has that schema, you are the citation source. If it does not, the AI guesses from whatever unstructured text it can find.
Google restricted the FAQ rich result (the accordion widget in search results) to government and health sites in August 2023. The SEO industry read this as FAQ schema is dead and stopped implementing it. That was a mistake. Google never stopped reading the schema. And AI search engines never used the Google accordion in the first place. They read the schema directly from your HTML. The deprecation that killed a Google display feature made the same schema more valuable for a faster-growing channel.
Organization or LocalBusiness schema
Organization schema tells AI engines who you are. It includes your legal name, URL, logo, description, founding date, and links to your social profiles. LocalBusiness extends this with physical address, phone number, geographic coordinates, business hours, price range, and cuisine or service category.
This is your entity data. When an AI engine builds a profile of your business, it looks here first. Without it, the engine infers your identity from page text, which is unreliable. With it, there is no ambiguity.
The most common problem is not missing schema. It is stale schema. Data that was accurate when the site launched but was never updated after a move, rebrand, or phone number change. If your schema says you are at one address and your homepage says another, the AI has to reconcile conflicting information. It may pick the wrong one.
Article schema
Article schema tells AI engines that a page is a piece of content, not a product page or a service page. It includes the headline, author, publish date, modified date, and image. This matters because AI engines treat content pages and commercial pages differently. If your blog post has Article schema, the engine knows it is a citable source of information, not a sales page.
The publish date and modified date are especially important. AI engines prioritize recent content for time-sensitive queries. If your Article schema says you published in 2024 and last updated in July 2026, the engine knows the content is current. Without dates in the schema, the engine has to guess from the page text, and it often guesses wrong.
BreadcrumbList schema
BreadcrumbList schema maps your site hierarchy. It tells the AI engine that this article lives under the blog section, which lives under the main site, which belongs to this organization. This helps the engine understand the context of your content and how it relates to the rest of your site.
It is the lowest-impact schema type on this list, but it is also the easiest to add. Most CMS platforms generate it automatically.
Which AI engines read which schema
All four major AI search engines read structured data from your HTML. The specifics vary, but the general pattern is consistent.
Google AI Overviews reads structured data the same way Google Search does. Google has the most mature structured data processing pipeline of any search engine. It reads FAQPage, Organization, LocalBusiness, Article, and dozens of other types. Google's documentation on structured data is the most detailed of any search engine, and their Rich Results Test is the standard validation tool.
ChatGPT, through its GPTBot crawler, reads your HTML including structured data. OpenAI has not published detailed documentation on which schema types ChatGPT prioritizes, but retrieval-augmented generation systems like ChatGPT's process all available HTML, including JSON-LD script tags. FAQPage schema is particularly effective for ChatGPT because the question-and-answer format maps directly to how users query the model.
Perplexity reads structured data through PerplexityBot. Perplexity is the most transparent about its sources, showing numbered citations that link back to the original page. Pages with clean structured data are easier for Perplexity to cite accurately because the extraction is unambiguous.
Claude, through ClaudeBot, reads structured data the same way. Anthropic has not published specific guidance on schema prioritization, but the retrieval pipeline processes HTML including JSON-LD.
The key takeaway: you do not need different structured data for different AI engines. The same FAQPage and Organization schema that Google reads is the same schema ChatGPT, Perplexity, and Claude read. Implement it once, and every major AI search engine benefits.
How to add structured data to your site
Structured data goes in your HTML as JSON-LD. That is a script tag with a specific type attribute that contains structured JSON. Here is what it looks like:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Do you take reservations?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes, we take reservations for parties of up to 8. Book online or call us during business hours."
}
}
]
}
</script>The script tag type must be application/ld+json. A regular script tag without that type will be treated as JavaScript and will not work. You can place the tag anywhere in the HTML, head or body. Google and AI crawlers find it either way.
If you use WordPress
WordPress does not add structured data by default beyond basic theme-level markup. You have three options:
- Use Yoast SEO, which generates Organization, Article, and BreadcrumbList schema automatically and includes a FAQ block for FAQPage schema.
- Use Rank Math, which does the same with a slightly different interface.
- Add JSON-LD manually using a Custom HTML block or a header injection plugin.
If you use a page builder like Elementor, the Accordion widget does not generate FAQPage schema unless you use a dedicated FAQ widget or add the schema manually. Many site owners assume their accordion widget handles schema. It usually does not.
If you use Shopify
Shopify themes generate Product schema automatically for product pages. For FAQPage and Organization schema, you need to add JSON-LD manually in your theme files or use an app. The Organization schema should go in your theme header and apply site-wide.
If you use a custom site or framework
Add JSON-LD script tags to your page templates. For Next.js, React, Vue, or any JavaScript framework, render the script tag server-side so AI crawlers see it in the initial HTML response. Client-side rendered schema works for Google (which executes JavaScript) but may not work for all AI crawlers, which may not execute JavaScript.
How to test your structured data
Three tools, each with a different purpose.
Google's Rich Results Test at search.google.com/test/rich-results validates your schema against Google's rich result requirements. Paste your URL or raw HTML and it tells you whether your FAQPage, Organization, or Article schema is valid. This is the fastest check.
The Schema Markup Validator at validator.schema.org checks your schema against the full schema.org specification, not just Google's requirements. This catches issues Google's tool ignores, like missing recommended properties or incorrect nesting.
Google Search Console shows you which pages Google has actually indexed with structured data. Go to the Enhancements section and look for structured data reports. If your pages have valid schema and Google has crawled them, they will appear here.
One caveat: Search Console reports only appear after Google detects the schema and processes it. If you just added it, expect a delay of days to weeks depending on your crawl rate.
Five mistakes that make your structured data useless
These are the problems that show up most often. Each one silently kills your schema effectiveness.
**Schema that does not match visible content.** Your FAQPage schema says yes, we offer gluten-free options but your visible page says we care deeply about all dietary needs. That mismatch creates a trust conflict. Some crawlers skip you. Others pick one version and ignore the other. The schema text should match the visible text. Not word for word, but the meaning must be the same.
**Malformed JSON.** A single trailing comma, a missing bracket, or an unescaped quote invalidates the entire schema block. JSON is unforgiving. Run your JSON through a validator like jsonlint.com before publishing.
**Wrong script type.** Using a script tag with type text/json or application/json instead of application/ld+json. Only the ld+json type is recognized as structured data. This is the most common implementation error.
**Hidden questions.** Adding FAQPage questions in the schema that are not visible on the page. Google considers this spam behavior. AI engines may ignore the entire schema block. Every question in your schema must have a corresponding visible question on the page.
**Blocked crawlers.** Your robots.txt blocks GPTBot, ClaudeBot, or Google-Extended. You have perfect structured data, and no AI crawler can read it. This is the most damaging problem because it removes your entire site from the AI engine's index. No amount of schema quality overrides a robots.txt block. Check that AI crawlers are explicitly allowed.
How structured data fits into the bigger picture
Structured data is one signal among several that determine your AI search visibility. The foundation is crawlability: if AI crawlers cannot reach your pages, nothing else matters. After that comes entity data (Organization schema), then content structure (FAQPage schema, heading hierarchy), then the content itself.
The sites that score best on AI search visibility get all of these right. They allow AI crawlers in robots.txt. They have Organization or LocalBusiness schema with accurate, current data. They have FAQPage schema on pages with real question-based content. They publish content that directly answers questions their customers ask.
The sites that score worst usually fail at the foundation. Blocked crawlers, missing entity data, no FAQ schema, content that does not answer questions. Structured data alone will not fix a site that blocks AI crawlers. But if you have the foundation in place, structured data is the fastest way to move from invisible to cited.
The practical takeaway
If you have not added structured data for AI search yet, start with two things.
Add Organization or LocalBusiness schema to your homepage. Include your business name, address, phone number, URL, logo, and description. Make sure the data is current. This takes 20 minutes.
Add FAQPage schema to your most important pages. Write five questions your customers actually ask. Write direct answers. Put them on the page as visible headings and paragraphs. Add the JSON-LD. Test it with Google's Rich Results Test. This takes 30 minutes per page.
Those two changes put you ahead of most of your competitors. The competitive advantage is large because most local and niche business websites have zero structured data for AI search. You can be one of the few who does by the end of the day.
If you want to know exactly where your site stands on structured data and every other signal that affects AI search visibility, [run a free audit at parceit.com](https://parceit.com/audit). The audit crawls your site, checks your schema, evaluates your content structure, verifies your crawler access, and gives you a prioritized list of what to fix first.
Frequently asked questions
Does ChatGPT read structured data?
ChatGPT retrieves web pages through GPTBot and processes the HTML, including JSON-LD structured data. OpenAI has not published detailed documentation on schema prioritization, but retrieval-augmented generation systems process all available HTML. FAQPage schema is particularly effective because the question-and-answer format maps directly to how users query ChatGPT.
Is FAQPage schema still relevant in 2026?
Yes. Google restricted the FAQ rich result visual feature to government and health sites in 2023, but Google never stopped reading or indexing the schema. AI search engines like ChatGPT, Perplexity, and Claude read FAQPage schema directly from your HTML. The schema is more valuable now than it was before the Google restriction because AI search is a faster-growing channel.
What is the difference between structured data and schema markup?
They refer to the same thing in practice. Structured data is the broader concept: data organized in a predictable, machine-readable format. Schema markup is the specific implementation: structured data written using the vocabulary from schema.org, typically in JSON-LD format. When people say add schema to your site, they mean add JSON-LD structured data using schema.org types.
Do I need different structured data for each AI search engine?
No. The same JSON-LD structured data that Google reads is the same data ChatGPT, Perplexity, and Claude read. There is no AI-engine-specific schema. Implement it once using standard schema.org types, and every major search engine benefits.
How long does it take for structured data to affect AI search visibility?
For Google, structured data is processed when Googlebot crawls your page. You can see the results in Search Console within days to weeks depending on your crawl rate. For AI search engines, the timeline is less predictable because their crawling and indexing cycles are less transparent. In general, expect to see effects within two to four weeks of implementation, assuming your site is crawlable and your schema is valid.