Guide · 9 min read
Schema markup for AI search: what it does and what it does not
Schema markup for AI search is one of the most oversold topics in GEO. Structured data helps machines identify who you are and what you sell, but no vendor has said it gets you cited, and Google says AI Overviews need no special markup. This guide separates what is documented from what is assumed, lists the types worth having and shows how to check that yours is correct.

What structured data is, in one minute
Structured data is a block of code in your page that describes its content in a fixed vocabulary, usually the one maintained at schema.org. Instead of a crawler guessing that a number is a price or that a name is the author, the markup states it.
Google recommends the JSON-LD format when your setup allows it. JSON-LD sits in a script tag in the page and does not change what visitors see, which makes it easier to maintain than markup woven into the HTML.
Two things it is not. It is not a ranking switch, and it is not a way to tell a search engine something your page does not say. Both points matter more for AI search than for classic results.
What Google says about AI Overviews and AI Mode
Google is explicit. Its page on AI features and your website says there are no additional requirements to appear in AI Overviews or AI Mode, that you do not need new machine-readable files or markup, and that there is no special schema.org structured data you need to add.
To be eligible, a page has to be indexed and eligible to show in Search with a snippet. The same page lists ordinary SEO practices, and one of them is making sure your structured data matches the visible text on the page.
So the honest reading is this: schema markup for AI search on Google is the same structured data you would add for classic Search. It can help Google understand the page and qualify it for rich results. It is not a separate ticket into the AI answer. For the rest of the picture, see how to appear in AI Overviews.
FAQ and HowTo rich results: where they stand
Many older guides still promise FAQ dropdowns and step-by-step HowTo snippets. Both are gone from Google Search at the time of writing:
- HowTo. In August 2023 Google limited HowTo rich results to desktop, and on September 13, 2023 it stopped showing them on desktop too, according to its announcement on HowTo and FAQ changes.
- FAQ. The same 2023 announcement restricted FAQ rich results to well-known government and health sites. Google's documentation changelog then says the feature would no longer appear starting May 7, 2026, and the FAQ rich result documentation was removed on June 15, 2026.
- Other types. In 2025 Google phased out several more, including course info, estimated salary, learning video, special announcement, vehicle listing and practice problem.
Google said in 2023 that unused structured data does no harm and there is no need to remove it. Valid FAQPage markup that mirrors a real FAQ on the page can stay. Just do not add it expecting a visible result in Google.
Do ChatGPT, Claude, Perplexity and Copilot read schema?
This is where most claims online go beyond the evidence. Here is what we could verify, vendor by vendor:
- Microsoft. In March 2025, at SMX Munich, Bing's Fabrice Canel said schema markup helps Microsoft's language models understand content, as reported by Search Engine Land and Search Engine Roundtable. That is a public statement by a Microsoft employee, not a formal specification of how Copilot uses it.
- Google. AI Overviews and AI Mode draw on the Search index, so Google's structured data handling applies there. Google does not say schema raises your chances of being cited.
- OpenAI. We found no OpenAI documentation stating that ChatGPT reads schema.org markup when it answers. For products, OpenAI documents a separate product feed that merchants upload; it does not reference on-page schema.
- Anthropic, Perplexity and Gemini apps. We found no public documentation either way. Unknown.
One technical point is fair to make: JSON-LD is part of the HTML, so any assistant that fetches the raw page receives it. Whether it is parsed, weighted or ignored is not published. For schema markup for AI search outside Google, treat it as a clear, low-cost signal of facts that should already be in your text, not as a lever with a known effect.
Schema markup for AI search: the types worth having
If schema works at all for assistants, it works by making facts about your entity unambiguous: your name, what you are, where you are, what you sell and who writes for you. These are the types that do that job for most businesses:
- Organization, on the home page or About page, with sameAs links to your official profiles.
- LocalBusiness, or the most specific subtype such as Dentist, Restaurant, Hotel or LegalService, if customers visit you or you serve an area.
- Product with Offer, on product pages, with price, currency and availability that match the page.
- Article or BlogPosting, on editorial content, with a named author and dates.
- Person, on author and team pages, linked from articles.
- Service, for service businesses. Google has no rich result for it, but it describes your offer in the same vocabulary.
- SoftwareApplication, for apps and SaaS products.
- FAQPage, only where the page shows the questions and answers to visitors.
BreadcrumbList is also cheap and useful for describing site structure. Skip types that do not describe your page just because a plugin offers them.
What a good Organization block contains
Google's Organization documentation has no required properties, only recommended ones. A solid block for a typical company describes:
- name, alternateName if you are known by a short form, and legalName.
- url of your home page and logo pointing to an image file that represents you.
- description in one or two plain sentences that match your About page.
- sameAs with the URLs of your official profiles: LinkedIn, Crunchbase, Wikipedia or Wikidata if they exist, and review platforms where you have a listing.
- address, telephone, email and contactPoint, using the same details shown on the site.
- Identifiers where they apply, such as vatID, leiCode or duns.
sameAs is the property most worth getting right. It ties your site to the profiles that describe you elsewhere, which is the same consistency we discuss in brand mentions for AI. Only list profiles you control or that are clearly about you.
Examples by type of business
Each list below describes the properties of one JSON-LD block. The values are fictional and only show the shape.
A local clinic, using the Dentist subtype of LocalBusiness:
- @type Dentist, name Northside Dental (fictional), url of the home page, image of the practice.
- address with streetAddress, addressLocality, postalCode and addressCountry.
- telephone, geo with latitude and longitude, and openingHoursSpecification for each day.
- sameAs pointing to the clinic's Google Business Profile URL and social profiles.
Local businesses have more to gain from consistent data than most; see GEO for local businesses.
An online shop product page:
- @type Product with name, image, description, sku and brand.
- offers as an Offer with price, priceCurrency, availability and url.
- aggregateRating only if real customer ratings are visible on the page.
Product data also travels through feeds; more in GEO for ecommerce.
A blog article and its author:
- @type Article or BlogPosting with headline, image, datePublished and dateModified, with time zone.
- author as a Person with name only, no job title in the name, and url pointing to an author page.
- On the author page, a Person block with name, jobTitle, worksFor pointing to your Organization, and sameAs to the author's professional profiles.
A software product: SoftwareApplication with name, applicationCategory such as BusinessApplication, operatingSystem, and offers with a price, set to 0 if free. For Google's rich result you also need aggregateRating or review. See GEO for SaaS for the content side.
A service business: Service with name, serviceType, description, provider pointing to your Organization, and areaServed with the countries or cities you cover.
How to validate your markup
- Run the page through Google's Rich Results Test, the first stop for any schema markup for AI search review. It shows which rich result types the page qualifies for and any errors or warnings.
- Run it through the Schema Markup Validator at validator.schema.org. It checks all schema.org types, including those Google does not use for rich results, such as Service.
- Use URL Inspection in Search Console to confirm Google sees the markup on the rendered page, not only in your template.
- Watch the rich result reports in Search Console after deployment. They flag errors across the whole site.
- Compare the markup with the visible page by hand on a sample of pages. No tool checks whether your price or opening hours are still true.
If your markup is injected with JavaScript, test the rendered output. Not every crawler renders JavaScript, which is why schema markup for AI search is safer in the initial HTML, a point we cover under technical GEO.
Common mistakes
- Marking up what the page does not show. Google's policies say not to mark up content hidden from readers and that structured data must be a true representation of the page.
- Stale facts. A price, address or opening hours changed on the page but not in the JSON-LD, or the reverse.
- Self-serving review stars. Ratings that a LocalBusiness or Organization collects about itself on its own site do not qualify for review snippets in Google.
- FAQ markup on pages with no visible FAQ, added only for the old dropdown.
- Several plugins outputting duplicate or conflicting Organization blocks.
- Generic types such as Thing or WebPage only, when a specific subtype exists.
- sameAs links to profiles that are abandoned, unofficial or about another company with a similar name.
Where schema fits in GEO
In our audits, schema markup for AI search is a hygiene check: we confirm it exists, matches the page and identifies the brand consistently. It takes hours, not months, and it removes one source of doubt about who you are.
What moves recommendations is still the rest: crawlers that can reach your pages, covered in technical GEO, pages that answer real questions, covered in how to write content AI cites, and mentions on sites assistants trust. A free GEO audit checks all three.
Frequently asked questions
Do I need schema markup to appear in AI Overviews?
No. Google says there is no special schema.org structured data needed for AI Overviews or AI Mode. Pages need to be indexed and eligible for a snippet. Schema can still help Google understand the page.
Does ChatGPT use schema markup?
OpenAI has not published documentation saying ChatGPT reads schema.org markup when answering. JSON-LD is in the HTML any fetcher receives, but how it is used, if at all, is unknown. Microsoft is the only vendor whose staff have said publicly that schema helps its models.
Should I remove my FAQPage markup now that FAQ rich results are gone?
There is no need if it matches a visible FAQ. Google has said unused structured data causes no problems. Remove it where it describes questions that are not on the page.
Which schema type matters most for a brand?
Organization, or a LocalBusiness subtype for local companies, with accurate sameAs links. It is the block that ties your site to the rest of your presence online, and it is the first one we check in schema markup for AI search reviews.
Can schema markup hurt me?
Markup that misrepresents the page can lead Google to ignore it or apply a manual action for spammy structured data. Accurate markup has no known downside.


