Structured data test: AI Schema Doctor

What is structured data?

Structured data is a data layer that describes page content in a labeled format that search engines and AI systems can read without guessing. Two tools show whether it is set up correctly: the AI Schema Doctor on this page extracts the JSON-LD blocks at the URL you enter and counts the types, while Google's Rich Results Test checks the same markup at field level. The standard is based on the schema.org vocabulary, founded in 2011 by Google, Microsoft and Yahoo and joined by Yandex the same year. The vocabulary contains more than 800 types; details such as titles, prices, authors, addresses and ratings are marked up with these types.

Schema markup is the implementation side of the same layer: schema.org is the catalog that defines which types and properties exist, and markup is the code for those types added to a page. The terms are often used interchangeably; structured data is the name of the concept and schema markup is the code in practice. The problem the layer solves is ambiguity: in plain HTML, an engine has to guess from context whether a value like "4.8" is a rating or a price, while the same value written into the ratingValue field of the AggregateRating type is pinned to a single meaning.

Which schema types does each kind of site need?

The right type depends on the kind of site; the goal is not to pile schema onto every page but to add the type that matches the page's actual content. A company site starts with three core types: Organization, WebSite and BreadcrumbList. On top of that, an e-commerce site adds Product, a content site adds Article, a business with physical branches adds LocalBusiness, and a service company adds Service. The tool on this page builds its report on the same eight core types and flags each missing one.

Expectations for FAQPage changed in 2026: Google removed FAQ rich results in May 2026, having already ended HowTo results in 2023. Neither type gets a special display on the results page anymore; their contribution is limited to making Q&A and step structures machine-readable. The right approach is therefore to use FAQPage when the page has a real Q&A block, not in the hope of a rich result; marking up content that is not there only adds error risk.

Why is JSON-LD the recommended format?

Google's structured data documentation explicitly recommends JSON-LD out of the three supported formats (JSON-LD, Microdata, RDFa). The reason is ease of maintenance: JSON-LD sits in a single script block, does not spread across the HTML template and stays intact even when the template changes. Debugging is quick because errors are looked for in one place, and it is easy to generate dynamically in a CMS or on the server. Microdata and RDFa are still read, so there is no need to migrate existing markup overnight. A typical Organization block looks like this:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Ornek Firma",
  "url": "https://www.ornekfirma.com",
  "logo": "https://www.ornekfirma.com/logo.png",
  "sameAs": ["https://www.linkedin.com/company/ornekfirma"]
}
</script>

The block works in either the head or the body, and Google reads both locations with no functional difference. A page can hold more than one block, and types can be nested: embedding Offer and AggregateRating inside a Product block is a common and valid setup. On the development side, the healthy practice is to generate blocks centrally in the template engine and output the right schema for each page type, rather than writing them by hand one at a time.

Why is schema a prerequisite for rich results?

Rich results are displays that go beyond the standard blue link: star ratings, price and stock lines, breadcrumb paths and event dates all fall into this class. Google only considers a display if the relevant type is marked up on the page without errors; a page without schema can never be eligible for these views. Eligibility and display are different things: error-free schema opens the door, and Google decides which queries show the enhanced result.

The validation workflow has two steps. Before publishing, enter a URL or code snippet into the Rich Results Test; the report lists valid items and errors item by item. After publishing, Search Console opens a separate status report for every display type it recognizes and sends new issues as notifications. The click effect comes from a mechanical advantage: a result that wins an enhanced display takes up more space on the page and stands out visually from plain links; the schema itself does not raise rankings.

Does schema bring visibility in AI engines?

The measured effect falls short of expectations. A difference-in-differences study of 1,885 pages published by Ahrefs in May 2026 found that adding schema did not increase AI visibility; in the same analysis, AI Overviews citations for pages with added schema fell by 4.6%. In other words, the promise of "add schema and AI will cite you" does not match current data. At this layer, schema belongs in the technical hygiene category: it clarifies identity and causes no harm, but it does not earn citations on its own.

The Türkiye context does not change this balance. According to Google's announcement, AI Overviews have been running for Turkish queries since February 18, 2026. The layer that really decides visibility is access: GPTBot, ClaudeBot and PerplexityBot do not run JavaScript and only read the initial HTML response. If the content and schema are not in that response, the page is empty for these engines; so the order of work is accessible, server-rendered content first, then error-free markup.

What does the AI Schema Doctor scan, and what does it not measure?

The AI Schema Doctor fetches the HTML source of the URL you enter on the server side and extracts every JSON-LD block inside script tags. The analysis is per page; you can enter the homepage or a single inner page, and scanning requires no sign-up. The report has three layers:

1. Types found: every schema type on the page is listed with its count; nested types and types defined inside @graph are counted too.
2. Comparison with recommended types: which of the eight core types, from Organization to Service, are missing is shown along with the reason.
3. AI readiness score: the share of the eight recommended types present on the page is converted into a score out of 100; if four are present, the score is 50.

What it does not measure is just as clear: the tool reads only JSON-LD, so Microdata and RDFa markup is not counted even when valid. It does not check required properties at field level; that layer is the job of the Rich Results Test. It does not run JavaScript either, so blocks injected later via Tag Manager do not appear in the report. A block with broken syntax counts toward the block total but produces no types; a gap between the block count and the type count usually points to broken JSON.

What are the most common schema errors?

Error families lead to one of two outcomes: the block is either not read at all or gives false confidence. These are the six families that keep recurring in scans:

• Markup that contradicts visible content: putting a rating, review or discount in the schema that is not on the page violates Google's structured data policy and can lead to a manual action notice.
• Syntax errors: a missing comma, an unclosed quote or an unescaped character invalidates the entire block; a single character takes down the whole schema.
• Type and content mismatch: adding Product to a service page or LocalBusiness to a blog post creates the wrong expectations and triggers warnings in validation tools.
• Conflicting blocks: on WordPress installs, the theme, the SEO plugin and a separate schema plugin can output the same Organization data with different values; the fix is to reduce schema generation to a single source.
• Missing required fields: if a Product block has no offers field carrying the price, the Rich Results Test flags a missing item and eligibility for enhanced display drops.
• Neglect: a block is added once and forgotten, then a theme or plugin update silently breaks the schema; that is why a rescan is needed after every change.

What does structured data deliver, and what does it not guarantee?

It delivers two things, identity and eligibility: what the page is becomes clear at the machine layer, and the page becomes eligible for rich result displays. The three things it does not guarantee are just as clear: display remains at Google's discretion, rankings do not rise because of schema, and since Ahrefs' May 2026 measurement, AI citations are not something to expect from schema.

The roadmap differs by audience. A developer with code access sets up the Organization, WebSite and BreadcrumbList trio in JSON-LD at template level, adds Product or Article depending on the site type and tests the output before going live. A site owner who manages things through plugins leaves schema generation to a single plugin, switches off conflicting blocks and finds the remaining gaps with a scan. The final step is the same for both: inventory checks on this page, field accuracy in the Rich Results Test.