Insights · AI search17 November 2026 · 7 min read

What schema markup helps AI search find your business?

Structured data tells AI search what your page is about without guessing. Which schema.org types matter for a business and the rule that keeps them valid.

The schema markup that helps AI search find a business is a small set of schema.org types, written as JSON-LD in the page head: Organization or a more specific business type for who you are, Service for what you offer, WebSite and WebPage for the site itself, FAQPage for pages that answer questions, Article for guides, and BreadcrumbList for where a page sits. The one rule that matters more than the choice of types is that everything in the markup must also be visible on the page. Google says so in its structured data guidance, and assistants that cross-check will distrust a page that claims one thing in its markup and shows another.

Why structure matters to assistants

An AI assistant answering "who does executive search for fractional CFOs in London?" has to work out, from your page, that you are an organisation, that executive search is a service you provide, that fractional CFOs are a seat you place, and that London is where you are. Prose makes it infer. Structured data lets it read. Google's guidance for its AI features says that no special markup is required and that the usual search requirements apply, and it also says that if you use structured data it must follow the guidelines and be validated. That is the right way to think about it: structured data is not a trick for AI search, it is the standard way of stating facts that every reader of your page, human or software, can rely on.

The types that matter for a business

Organization

Who you are: name, web address, logo, contact point, the places you serve, and sameAs links to your LinkedIn and other profiles so an assistant can connect the dots. If you are a specific kind of business there is often a closer type, such as ProfessionalService or LocalBusiness, that adds address and opening hours.

Service

What you offer, one entry per service, with a name, a plain description, the provider pointing back to your Organization, and the page address. If you publish prices, Offer can carry them. If you do not, leave it out rather than invent.

WebSite and WebPage

The site and the page, with the page's name and description matching the title and description tag. This lets a tool confirm it is on the right page of the right site.

FAQPage

For pages with questions and answers, each question with its answer, exactly as shown on the page. Assistants are question-driven, and this is the type that maps directly onto how they work.

Article

For guides and articles: headline, dates, author, publisher and the main image. It tells a tool this is an explanatory piece and when it was written.

BreadcrumbList

Where the page sits in the site. Minor on its own, useful as a consistent signal of structure.

A real example

getoperators, a site Orqen built and tuned for AI search, carries Organization, WebSite, WebPage, Service and Offer markup on its home page. The Organization states the practice's name, contact point and the countries it serves. The Services are the seats it places, one each, and the Offers carry the day rates the site publishes openly. Nothing in the markup is absent from the visible page. That is the whole discipline: state in structured form what the page already says in words. The site's home page is the place to see it, by viewing the source and searching for application/ld+json.

How to add it

  1. Write the page first. Markup follows content, never the other way round.
  2. Put a single JSON-LD script in the head with an @graph containing the types above, linked by @id so the Service points at the Organization and the WebPage at the WebSite.
  3. Generate it from the same data that builds the page, so it cannot drift. We build our sites this way: the title, description and FAQ in the markup come from the same source as the visible text.
  4. Validate it with the schema validator and with Google's rich results test, and fix every error before publishing.
  5. Check it again whenever the page changes.

What to avoid

  • Marking up reviews, ratings or awards you do not display or cannot stand behind.
  • Adding prices in markup that are not on the page.
  • Copying a competitor's markup and changing the name.
  • Marking up every paragraph as a FAQ when the page is not a FAQ.

Where this sits in the bigger job

Structured data is one of five things that make a site recommendable by assistants, alongside crawler access, plain pages, answers to buyers' questions and consistent facts. Our guide to getting recommended by ChatGPT puts them in order, and llms.txt explained covers the plain-text index that complements the markup. Orqen builds all of it into the websites we design and run.

Frequently asked questions

Does schema markup directly improve AI search rankings?

Nobody outside the companies that run the assistants can say exactly how it is weighted. What is documented is that Google requires markup to match visible content and be valid, and that assistants read structured facts more reliably than prose. Treat it as removing ambiguity, not as a ranking lever.

JSON-LD or microdata?

JSON-LD, in a script in the head. It is the format Google recommends and the easiest to generate and validate.

Can I add markup to a site built on a platform like WordPress or Webflow?

Yes. Most platforms allow a custom script in the head, and several have plugins or fields for it. The rule is the same: the markup must match the visible page.

How do I know it is working?

Validate it, then watch what the assistants say about you over the following months. If their description of your business starts to match your own words, the structure is being read.

Talk to us about your site

If you want structured data added properly, generated from your pages and validated, book a call with us. We will look at what your site has today and say what it needs.

Tell us what your business repeats.

One short call. A clear plan and a fixed price before anything starts.