Structured data for AI: schema that works - Ighenatt Blog

Structured data has long been a tool for communicating the meaning of your content to search engines. With the rise of generative AI engines (Google AI Overviews, ChatGPT Search, Perplexity), that function has become even more relevant: schema markup is the language that LLMs use to understand, categorise and cite your content with precision.

Google explicitly confirms that structured data facilitates content understanding by its AI systems. It is not a guarantee of appearance, but it is a signal that reduces ambiguity and increases the probability that your content will be selected as a source.

Why structured data is fundamental for AI engines

A language model can read and understand free text, but schema markup provides it with explicit metadata that eliminates the need to infer. When your page includes Article schema with author, date and organization, the AI doesn’t need to deduce that data from the HTML: it has it in a processable format.

This difference matters because AI engines evaluate the reliability of sources before citing them. An article with clear authorship (declared in schema), a verifiable publication date and an identified editorial organization is more trustworthy to the model than an article without that metadata, even if the textual content is identical.

68% of rich results in Google come from pages with JSON-LD implemented. JSON-LD is Google’s recommended format for its cleanliness (it doesn’t mix with content HTML) and ease of maintenance. Microdata and RDFa remain valid, but JSON-LD dominates modern implementations.

The schema types that generative AI engines value most

Not all schema types have the same impact on AI engine visibility. According to BrightEdge analysis and our experience in generative AI optimisation, these are the types with the highest correlation:

The Article + FAQPage combination is the most effective for informational content. Article provides the editorial context that AI needs to evaluate reliability, and FAQPage structures responses in a directly extractable format.

Article schema: how to implement it correctly for GEO

Article schema is the foundation for any editorial content page. The fields that most influence evaluation by AI engines are:

• author: full name of the author with a link to their profile page. If the author has a verified profile on Google Scholar or LinkedIn, linking to it reinforces the expertise signal.
• datePublished and dateModified: content freshness is a selection factor for AI Overviews. A recently updated article has an advantage over one published years ago without modifications.
• publisher: the organization backing the content, with logo and URL. This connects the article to the domain’s reputation.
• headline: must match exactly the visible H1 on the page. Discrepancies between schema and visible content are a negative signal.

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Exact title from H1",
  "author": {
    "@type": "Person",
    "name": "Author Name",
    "url": "https://example.com/author/"
  },
  "datePublished": "2026-03-06",
  "dateModified": "2026-03-06",
  "publisher": {
    "@type": "Organization",
    "name": "Ighenatt",
    "logo": { "@type": "ImageObject", "url": "https://ighenatt.es/logo.png" }
  }
}

The dateModified field is particularly relevant: when you update an article with new data, updating this date signals freshness to both Google and the AI models that prioritise recent information.

FAQPage schema: the questions AI answers come from here

FAQPage schema has a direct relationship with AI Overviews because it structures information exactly as LLMs process it: in question-answer pairs.

When Google generates an AI Overview for a question-type query, it searches for sources that contain structured answers. If your page has FAQPage schema with questions relevant to that query, the AI can extract the answer directly from the schema and cite you as a source.

The rules for effective implementation are clear:

• Each schema question must be visible on the page as text. If the user cannot see the question, Google will consider the schema deceptive.
• Answers should be concise but complete. Between 40 and 150 words is the optimal range for citability.
• Questions should be phrased as a real user would phrase them. Tools like Google’s “People Also Ask” or Answer the Public help identify the most frequent formulations.

A common mistake is marking questions in a commercial section with FAQPage schema (“How much do your services cost?”). Google distinguishes between genuine informational questions and promotional questions, and penalises the latter.

HowTo schema: step-by-step instructions that AI cites directly

For instructional content, HowTo schema is the most relevant type. It structures each step with name, description, image (optional) and position, allowing AI engines to extract sequential instructions with precision.

An important technical aspect: each step (HowToStep) must be self-contained. The AI may cite a single step or a subset of steps, so each one needs to make sense on its own without depending on the context of previous steps.

The fields that maximise citability are:

• name: short title of the step (5-10 words)
• text: complete description of what to do
• url: link to the step’s anchor on the page (facilitates direct navigation)
• estimatedCost and totalTime: additional data that AI can include in its responses

HowTo schema works especially well for queries of the type “how to do X” or “steps for Y”, which are the query type where AI Overviews has the greatest presence.

To validate that your implementation is correct, use Google’s Rich Results Test. This validator checks both the JSON-LD syntax and consistency with Google’s guidelines for each schema type.

Frequently asked questions about structured data and GEO

Does schema markup guarantee I will appear in AI Overviews?

No. Schema facilitates content understanding by AI systems, but appearing in AI Overviews depends on multiple factors: relevance, authority, content quality and E-E-A-T signals. Schema is one signal among many, not a guarantee.

Can I use multiple schema types on the same page?

Yes. Google recommends implementing all schema types relevant to each page. A blog page can simultaneously have Article, FAQPage and BreadcrumbList. The most effective combination for informational content is Article + FAQPage.

Does Google verify schema before using it in AI responses?

Google validates that schema is technically correct and faithfully represents the page content. If it detects discrepancies between the schema and the actual content (for example, FAQ markup on a page without visible questions), it may ignore or penalise the schema.

Implementing structured data correctly is one of the technical actions with the highest return for AI engine visibility. If you need an audit of your structured data or want to implement a schema strategy focused on technical SEO and GEO, talk to our team to design an implementation tailored to your site.