How much does structured data (Schema) matter for GEO? With Turkish examples

Structured data is code that describes what a page is (a company, a product, a list of questions and answers) using a shared vocabulary that machines can read, schema.org; today it is mostly written in the JSON-LD format. For GEO the short answer is this: Google writes that structured data is not needed to appear in its AI features, and the largest study measuring its effect on AI citations found no meaningful increase. Even so, structured data that is written correctly and consistent with the visible content keeps its value for rich results and for your brand's identity being read clearly.

In this post we discuss not structured data itself but its place in GEO. We have not measured its effect ourselves; every piece of information below rests on Google's documentation or on studies that explain their method, and the sources were reread on 7 October 2026.

What does Google say?

Google's AI features and your website document says there are no additional requirements or special optimizations for appearing in AI Overviews or AI Mode. According to the same document, you do not need to create new machine-readable files, AI text files or special markup to appear in these features.

Google's AI optimization guide, published in May 2026 and last updated on 10 July 2026, is even clearer. The guide lists "overfocusing on structured data" among common misconceptions: it writes that structured data is not required for generative AI search and that there is no special schema.org markup you need to add. The same paragraph adds that it is a good idea to keep using structured data as part of general SEO work, because it helps with eligibility for rich results in Google Search.

One recommendation in the AI features document concerns structured data directly: Google recommends making sure your structured data matches the visible text on the page. This shows that the general rule in the next section also applies to AI features.

The golden rule: match the visible content

Google's general structured data guidelines require structured data to represent the page's actual content, and state plainly that content readers cannot see should not be marked up. According to the guidelines, the markup should carry information that is current, original and not misleading.

Breaking this rule has a cost. According to Google, a structured data issue can lead to a manual action. A manual action caused by structured data means the page loses its eligibility for rich results; the guidelines also note that this does not affect the page's ranking in Google web search.

There is one more limit: Google does not guarantee that structured data will appear in search results, even when the markup is done correctly. Google's algorithm decides which feature to show by looking at variables such as search history, location and device type.

In practice this rule means:

FAQ markup: from 2023 to 2026

FAQPage was for a long time one of the most popular structured data types, because it made the questions appear as an expandable list in the search result. This changed in two steps.

The first step came on 8 August 2023. According to the announcement on Google's Search Central blog, FAQ rich results would from then on be shown only for well-known, authoritative government and health sites, and would no longer be shown regularly for other sites. The same announcement said there was no need to remove the markup, and that unused structured data causes no problems in Search but also has no visible effect.

The second step came in 2026. According to Google's documentation updates page, a deprecation note was added to the FAQ rich result documentation on 8 May 2026: the feature has not appeared in Google Search since 7 May 2026. In June 2026 the FAQ rich result documentation was removed entirely.

So today FAQPage markup does not produce a rich result in Google Search for any site. This does not mean the markup is invalid: FAQPage is still a valid schema.org type and other systems can read it. But we know of no published documentation on how those systems use it.

Effect on AI citations: what is known?

It is often said that structured data increases citations in AI answers. Two studies that explain their method do not support this claim.

Ahrefs: 1,885 pages and a control group

The study Ahrefs published on 11 May 2026 compared 1,885 pages that added JSON-LD between August 2025 and March 2026 with 4,000 similar control pages. Citations were measured in AI Overviews, AI Mode and ChatGPT in the 30-day periods before and after the markup was added, and the difference was calculated with a matched difference-in-differences method.

Ahrefs' conclusion is that adding structured data brought no meaningful increase in citations on any surface. The study's limits are also stated in the post: the pages examined were already receiving more than 100 AI citations a month, all structured data types were assessed together, the markup was often added at the same time as other changes to the page, and a 30-day window may miss slower effects. In other words, this study says nothing about a page that is never cited.

searchVIU: JSON-LD in live reading

searchVIU's experiment published on 2 December 2025 published the prices of fictional products on a test page in different forms (visible HTML, content built with JavaScript, JSON-LD only, and others) and asked ChatGPT, Claude, Gemini, Perplexity and Google AI Mode. None of the five systems reading the page at that moment found the price that existed only in the JSON-LD.

This experiment also has a limit, and the authors state it themselves: it tested only the live reading of the page. It does not show whether structured data is used at earlier stages such as indexing or training.

Next step

Check your site item by item

The checklist gathers, with the reason for each, what is checked so that AI crawlers can read and quote a page.

Open the checklist

Turkish JSON-LD examples

The examples below were written for a fictional bicycle shop; the address, phone number, price and profile links are placeholders. JSON-LD is placed in the page's HTML inside a <script type="application/ld+json"> tag. Every value in every example must be the same as the information visible on the page.

Organization and sameAs

Organization describes the organisation that publishes the page. The sameAs field lists other profiles belonging to the organisation and helps reduce confusion with another company of the same name. List only profiles that genuinely belong to you.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Örnek Bisiklet",
  "url": "https://www.ornek.com.tr/",
  "logo": "https://www.ornek.com.tr/logo.png",
  "description": "İzmir'de şehir ve dağ bisikleti satan, servis ve yedek parça hizmeti veren mağaza.",
  "foundingDate": "2012",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "Örnek Caddesi No: 1",
    "addressLocality": "Konak",
    "addressRegion": "İzmir",
    "postalCode": "35000",
    "addressCountry": "TR"
  },
  "contactPoint": {
    "@type": "ContactPoint",
    "telephone": "+90-232-000-00-00",
    "contactType": "customer service",
    "availableLanguage": ["Turkish", "English"]
  },
  "sameAs": [
    "https://www.linkedin.com/company/ornek-bisiklet",
    "https://www.instagram.com/ornekbisiklet",
    "https://www.youtube.com/@ornekbisiklet"
  ]
}

Product

Product is used on pages that describe a single product. The price, currency and stock status are written inside Offer; they must be exactly the same as the price and stock information visible on the page, and must be updated together when a campaign ends.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Örnek Şehir 28 Bisiklet",
  "image": "https://www.ornek.com.tr/img/sehir-28.jpg",
  "description": "28 jant, 7 vites, alüminyum kadrolu şehir bisikleti.",
  "sku": "OB-SEHIR-28",
  "brand": {
    "@type": "Brand",
    "name": "Örnek Bisiklet"
  },
  "offers": {
    "@type": "Offer",
    "url": "https://www.ornek.com.tr/sehir-bisikletleri/sehir-28",
    "priceCurrency": "TRY",
    "price": "1000.00",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition"
  }
}

FAQPage

FAQPage lists the questions and answers on a page. We saw above that it no longer produces a rich result in Google Search; if you use it, the questions and answers must also be readable on the page exactly as written.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Bisiklet bakımı ne kadar sürer?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Standart bakım çoğunlukla aynı gün tamamlanır. Parça değişimi gereken onarımlar, parçanın stok durumuna göre birkaç iş günü sürebilir."
      }
    },
    {
      "@type": "Question",
      "name": "Online aldığım bisikleti iade edebilir miyim?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Kullanılmamış ve ambalajı açılmamış bisikletler, teslimden sonraki yasal cayma süresi içinde iade edilebilir."
      }
    }
  ]
}

Run the markup you wrote through a validator before publishing it. Broken JSON-LD can be worse than none at all: it looks as if something is on the page, but no system can read it.

What can't we prove?

We want to state the limits of what we know on this subject plainly:

• We cannot prove that structured data increases AI citations. The largest controlled study we could find found no increase on pages that were already being cited.
• Nor can we prove that it does nothing. That study did not measure pages that are never cited, individual structured data types, or effects longer than 30 days.
• We do not know how ChatGPT, Claude and Perplexity use structured data when indexing. These companies' crawler documentation does not explain it; the live reading experiment tested only reading at that moment.
• We have no measurement of our own. We have not measured the effect of structured data on how often brands are mentioned or cited as sources in AI answers.

Within these limits the reasonable reading is this: structured data should not be sold as a lever for AI visibility, but when it is written correctly and consistently with the visible content it does no harm, and it has known benefits such as rich results and identity clarity.

What should you do?

Putting priorities in the right order also sets the right amount of time to spend on structured data:

1. Access first. If search bots cannot reach your page, the markup no longer matters. To check your robots.txt and CDN settings, see our robots.txt guide.
2. Then the visible content. Core information such as product details, prices, address and contact details should be written as text on the page. The searchVIU experiment shows that information sitting only in the markup was not found in live reading.
3. Next, consistent structured data. Describe who you are with Organization and sameAs, and your products with Product; match every value to the information on the page.
4. Don't waste time on extra files. Google's guide writes that Google Search needs no special markup or AI text files; we covered the status of llms.txt in a separate post.

You can use our checklist to check these steps item by item. Our scan, which needs no sign-up, also checks whether your page has valid JSON-LD and whether you give sameAs links to the external profiles that describe who you are.

Frequently asked questions

Is structured data required to appear in AI Overviews?

According to Google's documentation, no. Google writes that structured data is not required for AI Overviews and AI Mode and that there is no special schema.org markup you need to add. To be a supporting link, the page needs to be indexed and eligible to be shown with a preview in Google Search.

If I add structured data, will I be cited more in AI answers?

We found no evidence that shows this. Ahrefs' study comparing 1,885 pages with a control group found that adding JSON-LD brought no meaningful increase on pages that were already being cited. Because the study did not measure pages that are never cited, it says nothing about those pages.

Should I remove FAQPage markup?

You do not have to remove it. In 2023 Google wrote that unused structured data causes no problems in Search; FAQ rich results, meanwhile, have not been shown in Google Search at all since 7 May 2026. If you keep the markup, make sure the questions and answers also appear on the page exactly as written.

Which structured data types should I start with?

For most brands Organization is a good start: it gathers your name, address, logo and, through sameAs, the profiles that belong to you in one place. For sites that sell products, Product describes price and stock information consistently with the visible content. In every case the rule is the same: everything written in the markup must also be visible on the page.