Does structured data improve my position in Google?

Not directly. Google has repeated for years that markup is not a ranking factor and implementing it alone does not move a site up. It does help Google understand content unambiguously and opens the door to rich results, such as breadcrumbs or product details, and a richer result tends to collect more clicks.

Is structured data required to appear in AI Overviews?

No. In its May 2026 guide Google says directly that no special files and no special schema.org markup are needed for the generative features of search, and that optimising for AI is simply good SEO. Structured data remains, however, the most reliable way to hand machines the facts about your business.

Which structured data format should I choose?

JSON-LD, the format recommended by Google. It is a small block written into the page code: invisible to visitors, it changes nothing about the look or speed of the site, and it is easy to maintain and check with a validator.

Does an FAQ section still make sense now that Google retired FAQ rich results?

Yes. Since 7 May 2026 Google no longer shows expanded FAQ in search results, but the FAQPage markup remains valid and still helps machines understand the page. Above all, the questions and answers themselves are read by people and by AI models, because they answer real customer questions. The content keeps its value; only the decoration is gone.

How is structured data different from the llms.txt file?

Structured data is facts written into the code of each page, a standard developed since 2011 and read by search engines for years. llms.txt is a young proposal to summarise the whole site in a single text file for AI models. They complement each other rather than replace each other: first describe the pages properly, then summarise.

How do I check whether my site has structured data?

Paste your address into Google's free rich results test or into the schema.org validator. You will see exactly what data the machines read and whether there are errors. We can also check it for you and show the result on the specifics.

Structured data: how Google and AI actually read your website

← Back to blog Structured data schema.org - the site's layer of facts for Google and AI

Updated: 2 July 2026 - 7 min read

Your website has two audiences. The first is people: they see the layout, the photos, and they read the copy. The second is machines: Google's crawler, a screen reader, an AI model. That second audience sees nothing - it reads code. Structured data is the way to hand it the facts directly in that code, instead of making it guess. In 2026 this layer matters more than ever, though for different reasons than most guides promise.

We explain in plain words what structured data is, what it really does in Google, what Google itself says about it in the context of AI, and why the agentic version of the web is being built precisely on it.

What structured data is

Structured data is a short description of a page's content in a format that machines understand unambiguously. On a contact page a human sees the words "Call us: 500 600 700". A bot sees a string of digits that could just as well be a phone number, a price or a bank account. Structured data settles it: this is the phone number, this is the company name, this is the address, and these are the opening hours.

The shared vocabulary for these descriptions is called schema.org, and it is nothing niche: Google, Microsoft and Yahoo announced it together back in 2011, precisely so the whole industry would describe pages the same way. The vocabulary has types for most things a business wants to communicate: an organisation, a local business, an article, a product, a review, an event, a question and an answer.

Technically the most convenient format is JSON-LD, the one Google recommends: a small block written into the page's code, invisible to the eye, changing nothing about the look and slowing nothing down. Simplified, it looks like this:

{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "Your company",
  "description": "What you do, in one sentence.",
  "telephone": "+48 500 600 700",
  "address": { "@type": "PostalAddress", "addressLocality": "Raciborz" },
  "openingHours": "Mo-Fr 08:00-16:00"
}

A simplified example of JSON-LD with company facts. The machine no longer has to guess anything.

What it does in Google, and what it does not

Let us start with the disappointment worth knowing up front: structured data by itself does not lift your position in search. Google has repeated this for years - markup is not a ranking factor, and "we implemented schema" alone will not move a site up. If someone sells it as a ranking trick, they are selling a fairy tale.

The real benefits are two. First: Google understands the content unambiguously, so it confuses your business with something else less often and matches the site to the right queries more easily. Second: rich results in search - breadcrumbs instead of a raw address, and for shops the price, availability and product ratings, for events the dates, for job postings the details of the role. That is not a higher position, but a richer, more concrete result that tends to collect more clicks.

And here is an important 2026 update, because the list of those decorations is shrinking. Google retired HowTo rich results back in 2023. FAQ was first restricted to well-known government and health sites (August 2023), and on 7 May 2026 it stopped being shown altogether. The FAQPage markup itself remains valid and Google still reads it to understand pages better - it just no longer decorates the results. Anyone promising a client "stars and expandable questions in Google" for merely adding markup is promising the past.

What about AI: the facts from 2026

With the classic decorations fading, attention has shifted to one question: does structured data help you get into AI answers? Here we have the rare luxury of a first-hand answer. In May 2026 Google published an official guide to optimising sites for the generative features of its search engine, meaning AI Overviews and AI Mode. The position is clear: you do not need to add any special files or any special schema.org markup to appear in them, and the whole business of "optimising for AI" is simply good SEO. We wrote the same thing about the llms.txt file: a magic AI visibility switch does not exist.

Does that close the topic? No, because something more interesting is happening in parallel. The agentic part of the web - tools meant to find, compare or arrange something on a user's behalf - needs facts in a form it can take without guessing. And it is being built precisely on structured data. A good example is NLWeb, an open project from Microsoft that turns a website into a conversational interface for people and AI agents. As fuel it uses what a site already publishes: schema.org data and RSS feeds. In early 2026 Cloudflare added ready-made support, and early adopters include Shopify, Tripadvisor, O'Reilly and Eventbrite.

So the honest picture looks like this: structured data is not a ticket into AI Overviews, but it is the layer of facts that more and more machines reach for - from search engines, through screen readers, to AI agents. Whoever has this layer done properly does not have to chase every new fashion separately.

How to do it well

A good structured data implementation follows a few simple rules:

Markup must match the visible content. That is Google's requirement: you describe what is really on the page. Markup with content the page does not have, such as invented reviews, is a straight road to losing rich results and a manual penalty.
Describe what makes sense for your business. A few types are usually enough: company details, breadcrumbs, blog articles, questions and answers, and products in a shop. The point is not to mark up everything, but the facts that matter.
Check that it works. Google offers a free rich results test and schema.org has its own validator: paste the address and you see exactly what the machines read.
Update it together with the content. Phone number or opening hours changed? The markup has to change too, otherwise machines will keep repeating the outdated version.

What we do at Studio Kozubek

On the sites we build and run, structured data is part of the standard, not an add-on: company details, breadcrumbs, articles, question and answer sections - matched to what the site actually contains and maintained together with the content. This post has its own markup too (BlogPosting and FAQPage), even though Google no longer shows expanded FAQ results. We do it for understanding, not decoration.

In our approach it is the third link in the whole we describe in this series: a technically fast and stable site (that was the post about Agentic Browsing in PageSpeed), facts stated directly in the content and in structured data (this post), and a tidy summary in the llms.txt file. All three work towards one thing: that people, search engines and AI models understand the site without guessing. This is the foundation of our clients' SEO and websites, and we add AI integrations where they genuinely help - like the AI chat that answers customers on this site.

In short

Structured data is not a fashion and not a trick: it has been the same quiet standard for handing machines the facts since 2011. By itself it will not lift rankings and will not guarantee AI citations, but without it a site forces machines to guess, and guessing gets things wrong. If you do not know what your site tells machines, get in touch. We will check it on the specifics, with no scare tactics and no promised miracles.