Knowledge Graph Schema for AI: How to Speak JSON-LD to LLMs
If you are not speaking JSON-LD, you are speaking gibberish to an LLM. Learn how to map your brand as an explicit entity so Google and Perplexity understand exactly who you are.
Quick Answer
A knowledge graph schema is Schema.org structured data, written in JSON-LD, that defines your brand as an explicit entity — an Organization, its founder, and the topics it knows about — so large language models can identify and cite you accurately. Google recommends JSON-LD as its structured data format, and properties like knowsAbout and sameAs tell AI engines precisely who you are.
In the age of AI search, your website is no longer just a visual interface for humans. It is a data source for machines. If you want to control how your brand appears in an AI answer, you must speak the machine's native language: Schema.org JSON-LD.
Most brands treat Schema as an SEO checklist item—install a plugin, check a box, and forget it. That is a mistake in 2026. The Knowledge Graph is how a search engine reasons about the world, and it is only as good as the data you feed it.
What is a knowledge graph, and why does it decide whether AI cites you?
Google's Knowledge Graph is a model of real-world entities and their relationships—"things, not strings," as Google put it when it launched the graph in 2012.[1] An entity is a person, an organization, a place, or a concept. When an LLM or an AI Overview answers a question, it leans on these disambiguated entities to decide who is a credible source.
If your brand is not defined as a clear entity, the model has to infer who you are by crawling thousands of pages. That inference is noisy, and noisy entities get left out of answers to avoid hallucination. Structured data removes the guesswork.
Person vs. Organization: fixing entity ambiguity
The most common problem we see in audits is entity ambiguity. A consultancy is named after its founder. The site has a "Person" schema for the founder and an "Organization" schema for the company, but nothing links them. The AI is left asking: "Is this a person writing a blog, or a company selling services?"
You resolve this by declaring the relationship explicitly with the founder property on the Organization (and, where relevant, worksFor on the Person). You are building a machine-readable model of your real corporate structure—not decorating it.
Visual: The Schema Entity Hierarchy
Figure 1: Linking Organization, Founder, and Service into one graph removes the ambiguity that makes AI omit you.
What does a correct Organization schema look like?
Do not rely on the default output of a generic SEO plugin. Google supports JSON-LD, Microdata, and RDFa, but recommends JSON-LD as the easiest format to implement and maintain at scale.[2] Adding Organization markup to your homepage helps Google understand your administrative details and disambiguate your brand in search.[3]
Below is the actual Organization schema for this site. There is nothing hypothetical about it: the legal entity, registry code, and address are the same ones on our About and legal pages. Copy the shape, then swap in your own verified details.
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Vibe Marketing",
"legalName": "Tonotaco OÜ",
"url": "https://vibe-marketing.ai",
"logo": "https://vibe-marketing.ai/logo.png",
"description": "AI-marketing agency for Generative Engine Optimization (GEO), Answer Engine Optimization (AEO) and SEO. A division of Tonotaco OÜ, Tallinn, Estonia.",
"identifier": {
"@type": "PropertyValue",
"propertyID": "Estonian Business Registry Code",
"value": "16914587"
},
"address": {
"@type": "PostalAddress",
"streetAddress": "Ahtri tn 12",
"addressLocality": "Tallinn",
"postalCode": "10151",
"addressCountry": "EE"
},
"founder": {
"@type": "Person",
"name": "Tolga Guneysel",
"jobTitle": "Founder and Editorial Lead"
},
"knowsAbout": [
"Generative Engine Optimization",
"Answer Engine Optimization",
"Search Engine Optimization",
"AI-driven search"
]
}
Notice what is not here: no phone number we do not publish, and no sameAs profiles we cannot verify. Only claim what is true. A knowledge graph built on invented facts is worse than no graph at all, because a contradiction between your markup and the rest of the web tells the model your data is unreliable.
Why knowsAbout is the most important property for GEO
Schema.org defines knowsAbout as "a topic that is known about—suggesting possible expertise," and it applies to both Person and Organization entities.[4] It is the closest thing you have to handing the engine your résumé.
If you leave expertise implicit, the AI has to derive it from crawling your whole site. If you declare it, you steer retrieval. Be specific: do not write "Marketing." Write "Generative Engine Optimization" or "Programmatic SEO for SaaS." Match these topics to the concepts you actually want to be cited for.
How does sameAs connect your entity across the web?
Schema.org defines sameAs as the "URL of a reference Web page that unambiguously indicates the item's identity"—for example a Wikipedia page, a Wikidata entry, or an official profile.[5] It is how you tell an engine that the entity on your site is the same entity as the one on LinkedIn, Crunchbase, or Wikidata.
This merges otherwise-separate datasets into one stronger node. When Perplexity reads your LinkedIn and then your site, sameAs lets it fuse both into a single, higher-confidence entity instead of two weak, unrelated ones.
Critical rule: only list profiles you actually control and can verify. The snippet below uses an obviously generic placeholder company—replace every value with your own real, verified URLs.
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Example Corp",
"url": "https://example.com",
"sameAs": [
"https://www.linkedin.com/company/your-company",
"https://www.crunchbase.com/organization/your-company",
"https://www.wikidata.org/wiki/Q00000000"
]
}
Will FAQ or HowTo schema get you rich results?
No—and any guide that promises otherwise is out of date. In August 2023 Google restricted FAQ rich results to well-known, authoritative government and health sites, and later removed HowTo rich results entirely.[6] For a normal brand site, FAQ markup will not produce a rich snippet in the classic search results.
So why keep it? Because structured FAQ content is highly parseable for LLMs and AI answer engines. Use FAQ markup to feed clean question-and-answer pairs into the answer layer (GEO), not to chase a SERP snippet that no longer exists for you.
How do you validate your knowledge graph?
You cannot "set and forget" structured data. Validate it. Google's Rich Results Test checks eligibility for its features, while the Schema.org Validator is stricter about the markup itself. Run both, and re-run them whenever you change your entity data.
A perfectly validated knowledge graph is not a "verified badge"—there is no such thing to buy on Google—but it is the technical proof that you are who you say you are. Consistency across your site, your markup, and your third-party profiles is the signal that earns trust.
Frequently asked questions
What is a knowledge graph schema?
It is Schema.org structured data—usually JSON-LD—that describes your brand and its people, services, and topics as linked entities. It gives search engines and LLMs a machine-readable definition of who you are instead of leaving them to guess from page text.
Is JSON-LD better than Microdata for AI search?
Google supports JSON-LD, Microdata, and RDFa equally, but recommends JSON-LD because it is the easiest to implement and maintain at scale.[2] A single script block in your <head> is simpler to keep accurate than markup interleaved through your HTML.
What is the knowsAbout property used for?
It declares the topics an entity has expertise in. Schema.org lets you use it on both Person and Organization, with plain text, a Thing, or a URL as the value.[4] Use it to state the specific concepts you want AI engines to retrieve you for.
Does the sameAs property require a Wikipedia page?
No. Wikipedia and Wikidata are strong references, but Schema.org accepts any reference page that unambiguously identifies your entity—including your official profiles.[5] Only list URLs you control and can verify; never invent one.
Will FAQ schema give me rich results in Google?
For most sites, no. Since August 2023 Google shows FAQ rich results only for authoritative government and health sites, and HowTo rich results were removed entirely.[6] Keep FAQ markup for AI-answer citation, not for classic SERP snippets.
Should I use Person or Organization schema for a personal brand?
Use both, and link them. Declare the Organization, declare the Person, and connect them with founder and worksFor. That resolves the ambiguity that makes an AI unsure whether you are a company or an individual.
How do I check that my schema is valid?
Run it through the Schema.org Validator for markup correctness and Google's Rich Results Test for feature eligibility. Re-validate after every change to your entity data so your markup never drifts out of sync with reality.
Action Item:
Audit your homepage schema today. If it is just default plugin output, define a real Organization entity, link your founder, fill knowsAbout with the concepts you want to own, and add only sameAs profiles you can verify. Then validate and deploy.