Structured data

What schema markup do I need for AI search?

Structured data is where a lot of effort gets wasted on the wrong things. Four types do almost all the work, and one detail matters more than all of them.

The short answer

You need Organization, WebSite, Service and FAQPage, and each one needs a stable @id. The identifier matters more than the volume: without it, every page declares a brand new anonymous business and nothing accumulates.

The detail almost everyone misses

Most sites emit correct, valid schema that still fails to do its job, because every page mints a fresh entity.

Schema uses an @id field as a stable identifier. When your homepage says your organisation is a nameless object, and your about page says the same, and your service page says it again, a machine sees three separate assertions with no way to know they describe one business. Authority spreads thin across all three and compounds on none.

Give the organisation one permanent id, use the same id on every page, and reference it by id everywhere a page needs to mention the business. Now every page is evidence about one thing. This is a small change with an outsized effect and it is the first thing worth fixing.

The four types that matter

Getting sameAs right

The sameAs field lists other places on the web that are the same entity. It is how you tell a machine that the company on your site, the company on your professional profile and the company in a public register are one business.

Include a public register entry where your jurisdiction has one, your main professional and social profiles, and any authoritative listing you control. Keep it to profiles that genuinely represent the business and that you keep current. Ten links to abandoned profiles are worse than three to live ones, because inconsistency is the thing you are trying to remove.

What to skip

Do not add schema types that do not describe you. Marking a page up as a Product because Product markup sounds valuable produces contradictions, and contradictions cost more than the missing markup would have.

Do not add aggregate ratings you cannot evidence. Inventing a rating is the fastest way to become a source a system stops trusting, and it is the kind of error that is trivially caught.

Do not chase breadth. Four correct, connected, consistently identified types beat fifteen disconnected ones every time.

And do not treat schema as the whole job. It describes what is on the page. It cannot invent an answer that the page does not contain.

Checking it actually works

Validate the markup, then do the harder check: fetch the page as a crawler and confirm the structured data is present in the HTML that comes back rather than being injected later by JavaScript. Markup that only exists after a browser runs your scripts is markup a retrieval system may never see.

Then check the ids resolve to one entity across the site. View a few different pages and confirm they all reference the same organisation identifier rather than each declaring a new one.

Published 07 September 2026 by Carpe DM Strategies. We run this work on our own business every day, which is where the examples come from.

What schema does AI search actually use?
Organization, WebSite, Service and FAQPage do almost all the useful work. They tell a system who you are, what you sell, and which questions you answer, in a form that needs no interpretation.
Why does my schema need an @id?
Without a stable @id, every page declares a brand new anonymous organisation. A machine cannot merge those into one business, so authority scatters across many weak entities instead of accumulating on one strong one.
Is FAQ schema still worth adding?
Yes. Even where search engines have reduced how often they display FAQ rich results, the markup remains one of the most directly extractable structures available to an AI retrieval system.
Can schema markup alone get me cited by AI?
No. Schema describes what is on the page. If the page does not actually answer the question, there is nothing worth citing, and no amount of markup creates it.
Should schema be JSON-LD or inline markup?
JSON-LD in the page head. It is the format the major engines recommend, it is easier to keep correct, and it is simpler to confirm it is present in the raw HTML.

Want your entity wired up properly?

Book a free 30-minute call. We will show you what your current markup declares and what it should declare instead.

Book a Free Call