Schema Markup for AI Search and What It Does and Does Not Do

Schema markup is a set of labels you add to a page so machines can tell what the page describes.

For AI search, Google states that no special schema markup is needed to appear in AI Overviews or Google AI Mode. Schema still helps search engines identify your organization, pages and products, and that is where its value sits.

This guide is for owners and marketers who keep being told that structured data is the secret to AI visibility.

It covers three things, and it separates them: what Google documents, what schema reliably does, and what is only assumed. It ends with a short, safe way to add and check markup on a small site.

The short answer: add accurate schema because it makes your pages easier to identify and can earn rich results, not because anyone has shown it makes an AI engine quote you. Spend the larger share of your effort on the content and the technical basics that Google says matter most.

Schema markup labels what the page already says

Search engines read your words and work out their meaning, and AI generated summaries are built on the same understanding. Schema markup removes some of the guesswork by labeling the parts.

Google describes structured data as a standardized format for providing information about a page and classifying its content, and gives a recipe page as the example: the ingredients, the cooking time and the calories each get a label (Google Search Central).

The vocabulary comes from schema.org, a shared set of types and properties. Schema.org says its vocabularies are built so that pages can be understood by the major search engines, naming Google, Microsoft, Yandex and Yahoo (Schema.org).

A page about a lodge, for example, can say that it describes a LodgingBusiness with a name, an address and a phone number. The words on the page stay the same. Only the labels are new.

Google supports three formats: JSON-LD, Microdata and RDFa. It recommends JSON-LD, a block of code placed in a script tag that sits apart from the visible text, which makes it easier to maintain (Google Search Central).

Here is a small illustration of Organization schema in JSON-LD, with an invented business:

“` <script type=”application/ld+json”> { “@context”: “https://schema.org”, “@type”: “Organization”, “name”: “Example Lodge”, “url”: “https://www.example.com”, “logo”: “https://www.example.com/images/logo.png”, “sameAs”: [“https://www.linkedin.com/company/example-lodge”] } </script> “`

Google states that AI answers need no special schema

This is the most important sentence in the whole debate, so it is worth quoting from the source and not from a summary.

Google’s guide to AI features says: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.” (Google Search Central).

Google’s guide to generative AI search repeats the point in its list of things you can ignore. It says structured data isn’t required for generative AI search, and it still calls structured data a good idea as part of overall SEO, because it helps with eligibility for rich results (Google Search Central).

The same guide explains how the AI features work. They use retrieval-augmented generation, which grounds AI models in pages found by core Search ranking systems, and query fan-out, which runs several related searches at once.

Eligibility is plain: a page must be indexed and eligible to show with a snippet, and there are no additional technical requirements.

So the supported position is narrow and clear. Schema is not a requirement for Google’s AI answers and has no special AI version. These statements describe Google’s own features only. They do not describe how every other AI assistant or AI engine works.

AI crawlers must reach the page before they can read any markup

Schema only works if the page can be fetched, and that is true for every one of the AI crawlers as well as Googlebot.

Google tells site owners to make sure crawling is allowed in robots.txt and by any CDN or hosting infrastructure, and its structured data guidelines say not to block pages that carry markup with robots.txt, noindex or any other access control (Google Search Central).

The same logic applies to other AI search engines. OpenAI’s crawler overview says sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, and it recommends allowing OAI-SearchBot in robots.txt. A robots.txt change can take about 24 hours to take effect (OpenAI).

That page describes access, user agents and IP ranges. It does not mention schema markup or structured data. Access is the documented lever for ChatGPT search, and markup is not. Whether AI systems read the markup is a separate question from whether they can fetch the page.

What schema markup reliably does

Schema has real, documented effects. They just are not the ones sold in many AI Overviews and AI visibility pitches.

It can make a page eligible for rich results. Google says structured data can enable more engaging search results, called rich results. It lists case studies from the sites themselves, and for more detail you can read them on Google’s page: Rotten Tomatoes measured a 25% higher click-through rate on pages with structured data, and Nestlé measured an 82% higher click-through rate for pages shown as rich results (Google Search Central).

Those are the sites’ own measurements about classic rich results, not AI answers, and they come from large publishers. Treat them as proof that rich results can lift clicks, not as a forecast for your site.

It helps Google identify your organization. Google says Organization markup on your home page can help it understand your organization’s administrative details and disambiguate your organization in search results. Organization markup describes the business behind the entire site, which is why Google points it at the home page. Search engines and AI systems both have to decide which business a page is about. Some properties influence which logo appears in Search results and in your knowledge panel. There are no required properties (Google Search Central).

It gives Google structured facts about the world. Google says it uses structured data to understand the page and to gather information about the people, books or companies named in the markup. It also says it can make general use of the sameAs property, and that some elements may be used to enable future Search features.

That last point is the honest way to read the AI claim. Clear labels may help AI systems that read them.

Google has not said they are used to choose AI Overview citations or any other AI citations of pages. None of the Google pages we opened describes structured data as a ranking signal for AI answers.

What schema markup does not do

Four limits come straight from Google’s documentation.

It does not guarantee anything. Google says using structured data enables a feature to be present but does not guarantee that it will be. A page can be marked up correctly and still show as plain text, with no rich result and no AI Overview visibility gain, because the algorithm tailors results to the person searching (Google Search Central).

It does not replace good content. Google says creating content that people find unique, compelling and useful will likely influence your presence in generative AI search more than any other suggestion in its guide. It warns against overfocusing on structured data.

It does not unlock a hidden channel. Google’s list of things to ignore also covers llms.txt files, chunking content and rewriting content just for AI systems. Schema belongs on the same list when it is treated as a trick. It is part of traditional SEO, not a replacement for it.

Some of it has already been switched off. Google’s documentation changelog says the FAQ rich result stopped appearing in Google Search from May 7, 2026, and that its documentation for the feature was removed (Google Search Central). Advice to add FAQ schema for visibility is out of date for Google’s own results.

That does not make your FAQ section worthless. Readers still use it. It only means the markup behind it is no longer a rich result feature.

AI engines citing you because of schema is assumed, not shown

Many guides state that schema makes AI engines cite you and earns AI citations, or that FAQ schema feeds AI answers. Two things are missing from those claims: a statement from the search company, and a controlled test that isolates the markup.

A controlled test would change only the schema on a page and measure citations before and after, across enough pages and queries to rule out chance. We did not find one from an original publisher, so we leave such figures out. The same goes for percentages that appear only in vendor roundups.

That leaves four claims you will meet. Here is how each stands against the sources we opened. Check any claim against multiple sources, and treat the search company’s own documentation as the trusted source.

If you want proof for your own site, run a small test and measure it. The section on checking results below shows how.

An illustration of how schema can help AI understand one business

Many vendor guides add that AI search systems see the web as a network of connected entities. Google does not say this about its own AI features. The idea still explains why consistent naming matters. Take an invented business, Example Lodge.

Its home page calls it “Example Lodge”, its contact page says “Example Safari Camp”, and its profile on a listing site says “The Example Lodge and Spa”. A person can guess these are the same place. AI systems generating answers have to guess too, and schema’s job is to remove ambiguity.

Marking up the home page with Organization schema, and listing the official profiles with sameAs links, gives those systems an explicit signal about which pages and profiles describe the same entity, because the references point to one record.

Google’s guidelines say to use @id to link items that belong together on a page, such as a recipe and its video.

The same habit works across a site: give the Organization one identifier and refer to it from your Article markup, so every page points to the same entity. It only works if the visible text agrees. If the markup says one name and the page says another, you have created a new conflict.

This is a labeled illustration, not a case result. It shows the job schema is built to do, which is to reduce ambiguity, and it says nothing about whether any AI answer will name the lodge as a result.

Five schema types are enough for most small sites

Google’s gallery of supported structured data lists the features it can show, among them Article, Breadcrumb, Organization, Local business, Product, Review snippet and Video (Google Search Central). A small site does not need all of these schema types, and other schema types exist for events, jobs and recipes.

Most small sites can start with Organization schema and Article schema. Use the most specific type that applies. Google’s general guidelines say to use the most specific applicable schema.org type and property names, and to put the markup on the page it describes (Google Search Central).

Do not add structured markup for a type just because a plugin offers it. A service business with no product pages gains nothing from Product schema, and adding Product schema to a service page invites errors. Product schema belongs on product pages.

Add and check schema in five steps

The routine for schema markup for AI is the same as for any other search feature, and it takes one page type at a time. Start with the home page, then move to the blog.

1. Write the visible facts first

Schema must describe the visible content, meaning what readers can see. Google’s guidelines say not to mark up content that is not visible to readers of the page. If your phone number is not on the page, it does not belong in the markup.

2. Pick the most specific type

Choose from the types in the table above. Check schema.org for the exact type name, such as Organization, and fill in the properties that apply. Google advises fewer but complete and accurate properties over many half-filled ones.

3. Add JSON-LD to the page

Place the JSON-LD block, which is a single script block, in the page template.

On a CMS such as WordPress, Wix or Shopify, Google says you may not be able to edit the HTML directly, and that a plugin or search settings page can add the markup. Put the same markup on duplicate pages, not only the canonical page.

4. Validate before and after launch

Use Google’s Rich Results Test to check that the markup is valid and eligible, and the schema markup validator to check it against the full vocabulary. The schema markup validator checks syntax against schema.org, while the Rich Results Test checks Google eligibility.

Fix errors the tools flag, starting with the critical ones. Google says its own documentation is definitive for Google Search behavior, so when the two tools disagree, follow Google’s.

5. Watch Search Console, then judge the result

After launch, use the URL Inspection tool to see how Google reads the page. Google recommends before and after tests on a few stable pages, tracked in the Performance report in Search Console for several months.

For AI features, Google points to the Generative AI performance report in Search Console. Revalidate regularly, because Google says markup can break after deployment through templating or serving issues, which is why it points to the rich result status reports. Our Search Console entry explains the tool.

A small controlled experiment beats a big claim

If you still want to know whether schema moves anything for you, run a small controlled experiment. Google’s own advice for measuring structured data is a before and after test on a few pages, and the same idea works for AI answers.

Pick six pages of the same type with stable content. Add valid markup to three and leave three untouched. Keep every other change off the list for a few months. Record impressions and clicks from Search Console, and note by hand whether your pages are named in AI answers for a fixed set of questions.

Six pages will not prove a general law. They will tell you whether the effort paid for itself on your site, and that is the only answer you need. Our Search Console entry covers the reports to watch, and our self-check am I visible in ChatGPT gives you the fixed questions to track.

Common mistakes make good markup harmful

Most schema problems come from adding markup faster than the page changes. These are the ones Google’s guidelines name or imply.

• Marking up hidden content. The markup must match the visible content. A mismatch can make a page ineligible for rich results.
• Marking up fake or irrelevant content. Google warns against fake reviews and content unrelated to the page. Review markup is the easiest to abuse: Google says reviews or star ratings that are not by actual users may result in manual action, which removes rich result eligibility.
• Creating empty pages to hold markup. Google says not to create blank or empty pages just to hold structured data.
• Blocking the page. Do not block structured data pages from Googlebot with robots.txt, noindex or other access controls.
• Letting markup go stale. Provide up-to-date information, because time-sensitive content that is no longer relevant will not show as a rich result.

A manual action for structured data removes rich result eligibility. It does not change how the page ranks in web search, according to Google.

Schema works best when entity clarity is already in place

Schema labels things. It cannot make a vague business clear. If your pages describe your organization in three different ways, with different names, services and addresses, labels will only record the confusion.

Our Library describes the common fault: a site that names its organization differently from page to page splits its account of itself. The fix is editorial first.

Write one plain description, use one name and keep it the same across your site and profiles. Then mark it up. Schema helps most when the wording is already consistent.

Think of schema as the last coat, not the foundation. It supports search visibility and AI visibility, but it cannot make AI understand a business that describes itself three ways.

The foundation is the work Google puts first: unique content people find useful, a site search engines can crawl and index, and accurate business details. Our guide to a search-friendly website covers the technical basics, and the technical SEO entry defines them.

For the wider picture, read our pillar on what GEO is, where schema is one small part of being found in search and AI answers. Our briefing on how AI search changes SEO sets out what the evidence shows so far.

Content strategy comes before schema in AI search optimization

If you have limited time, work in this order. Each step has a documented reason, and schema comes after the steps Google puts first.

1. Publish content with a point of view that is not available everywhere else.
2. Confirm that important pages are crawlable, indexed and eligible to show with a snippet.
3. Check that your business details are accurate and consistent across your site and profiles.
4. Add Organization schema to the home page and add Article schema to each post, matching the visible page.
5. Measure in Search Console and adjust.

Schema sits inside a wider content strategy for AI driven search, and the order above follows the emphasis in Google’s own documents. Our guide to answer engine optimization and the classic guide to SEO go deeper on steps one to three.

The Brand Library defines generative engine optimization, and its entry on AI Overviews explains the feature Google keeps referring to. A citation is what you are hoping to earn, and nothing in the schema documentation promises one.

If you want a second pair of eyes on how your brand is found in search and AI answers, see our Distribution Strategy service.

Get the next step in your inbox

To see what the AI engines say about your business today, and what to change so they say something better, read our free ebook When AI Answers First. It pairs well with the checks in this article.

For one argument and one test a week, join Ground Truth, our letter on brand decisions, sent on Wednesdays.

FAQ