Does schema markup help you get cited by AI?

No markup makes a page eligible for an AI answer. Google states plainly that structured data is not required for generative AI search and that there is no special schema.org markup to add for it, while remaining worth using as part of ordinary SEO because it helps with eligibility for rich results.

So anyone selling schema as the route to being cited is selling something Google has publicly said does not exist.

What structured data does do is remove ambiguity. It tells a machine which string is a price, which is a service, and which company all of it belongs to. For a company a system has little prior information about — a new name, a new domain, a name that collides with something else — that is worth a great deal. Use it for that reason. Do not expect it to do the other thing.

What Google actually says

Google published a guide specifically about optimising for its generative AI features. Its central position is that these features are built on the same core Search ranking and quality systems as everything else, and use retrieval-augmented generation and query fan-out to surface material from the Search index. There is no separate track.

On markup, the guide is direct: structured data is not required for generative AI search, and no special schema.org markup exists for it. The guide also states that from Google Search's perspective, optimising for generative AI search is optimising for the search experience, and is therefore still SEO.

There is no secret schema. There never was one to find.

Source: Optimizing your website for generative AI features on Google Search, Google Search Central.

Then why do we still build it properly?

Because ambiguity is expensive, and we are our own worst case.

"VerityNodes" is a new string. Search results for it are dominated by blockchain material — verifier nodes, node lists, a token of a similar name. The existing machine association for our own name is an industry we have nothing to do with. Structured data is how we state, unambiguously and in the same words on every page, that this company is a consultancy measuring AI visibility for equipment manufacturers, registered in the United Kingdom, with one email address and one founder.

That does not make us eligible for anything. It removes a wrong answer. For a manufacturer whose name is shared with a town, a common noun or a larger company in another sector, the same logic applies and the stakes are higher.

What it actually does

Three things, all of them modest and none of them magic.

It states what things are. A page can say "from $4,500" in a sentence; markup says that this is a price, in this currency, for this named service, offered by this organisation. It ties everything to one identifier, so pages across a site describe one company rather than several similar ones. And on Google Search it makes pages eligible for rich results, which is its documented job and the reason it existed before any of this.

We also never mark up anything that is not visible on the page. Markup is a description of what is there, not an additional claim made to machines in private.

What it does not do

It does not make a page eligible for AI answers.

It does not substitute for corroboration. A page can be perfectly marked up and still be the only place on the internet making its claim, which is a weak position however well it is described.

And it does not rescue a page that fails to answer the question it was written for. Markup describes content. It cannot improve it.

The tactics that do nothing

Google's guide names several practices as unnecessary for its generative AI features, and they are worth listing because they are exactly what is being sold in this market: AI-specific files such as llms.txt, forced chunking of content, AI-targeted rewrites of otherwise good pages, chasing inauthentic mentions, and any special or invented schema for AI search.

If an agency's proposal leads with any of those, the proposal contradicts the published guidance of the company whose systems it claims to influence. That is a reasonable thing to ask them about.

Where the real work is

Two places, and neither is markup.

The first is deciding what a machine has to be able to verify about your company and making every source agree with it — what we call entity mapping, defined on Map. Markup is how that agreement is stated on your own pages. It is the notation, not the argument.

The second is covering your category well enough that you are relevant to the questions being asked at all, which is topical authority, defined on Build. A well-marked-up site covering three questions loses to a plainly-marked-up site covering thirty.

If you only do three things

Put an Organization block on every page, using one identifier for the company and never redefining it. Make the name, address, email and description identical everywhere they appear, including on directories and listings you do not control. Then validate what you have built, in Google's Rich Results Test and the Schema.org validator, before it goes live.

That is a day of work for most sites, it is learnable, and it is the kind of thing worth doing in-house — see can you do this yourself.

What this article cannot tell you

It cannot tell you that adding schema will change your position in AI answers. Nobody can isolate one change in a system with this many inputs, and any firm presenting markup as the cause of a movement is describing a correlation they did not control for.

It cannot promise the guidance is permanent. It was published in May 2026 and updated in July. We will update this page when it changes, and the date of that change will be on it.

And it cannot settle the wider argument. Plenty of practitioners report that markup correlates with being cited. That may be true without markup being the cause — a company with thorough structured data is usually a company doing everything else properly too.

What to do next

Build it properly, expect it to remove ambiguity rather than win citations, and spend the time you save on the two things above.

If the underlying question is how long any of this takes to show in the answers, that is set out in how long this takes to work. If it is whether AI systems currently name you at all, that is measurable, and the AI Visibility Audit is what measures it.