Google something like "Apple" and you see a panel on the right with the logo, founding date, CEO, stock price. That's the Knowledge Graph at work. It knows "Apple" the company is different from "apple" the fruit, and it knows a ton of facts about both.
What a knowledge graph actually is
A knowledge graph is essentially a giant database of things and how they relate to each other. Not just keywords and documents, but actual entities: companies, people, products, concepts, places.
Each entity has attributes (Apple was founded in 1976, its CEO is Tim Cook) and relationships to other entities (Apple makes the iPhone, Apple competes with Samsung, Apple was co-founded by Steve Jobs).
This structure lets systems answer complex questions. Not just "find pages with these keywords" but "what companies make smartphones" or "who founded Apple."
Why AI systems rely on knowledge graphs
Large language models are trained on text, which gives them general knowledge. But for current, factual, entity-specific information, they often tap into knowledge graph-like structures, either through their training data or through real-time retrieval.
When you ask ChatGPT about a specific company, it's drawing on structured knowledge about that entity. What category is it in? What does it do? How does it compare to similar entities?
If your brand exists clearly in these knowledge structures, AI can confidently talk about you. If you're a fuzzy, poorly-defined presence, AI has less to work with.
How to get into the knowledge graph
For Google's Knowledge Graph specifically:
Claim your Google Business Profile. This is the most direct way to establish your entity with Google.
Use structured data (schema markup). Organization schema, product schema, person schema. This explicitly tells search engines what kind of entity you are.
Wikipedia and Wikidata. These are major sources for knowledge graphs. If you're notable enough for a Wikipedia entry, that's a strong signal. (Don't create one yourself, that violates their rules. But you can ensure accurate information on Wikidata.)
Consistent information everywhere. Same name, same description, same key facts across your website, social profiles, directories. Knowledge graphs reconcile information from multiple sources. Consistency helps.
The connection to AI visibility
Here's why this matters for getting recommended by AI: knowledge graph presence signals legitimacy and clarity.
When AI tools decide who to recommend, they're essentially looking at entities and comparing them. If your entity is well-defined in these systems, if it's clear what you are, what category you're in, what your attributes are, you're easier to recommend.
A brand that exists as a clear entity with consistent information is easier to trust than one that's a vague collection of web pages.
Practical steps
Most small businesses won't get into Wikipedia, and that's fine. Focus on what you can control:
- Google Business Profile (if applicable)
- Schema markup on your website
- Consistent NAP (name, address, phone) across directories
- Clear, consistent brand description everywhere
- Profiles on relevant platforms that feed into knowledge systems (LinkedIn, Crunchbase for B2B, etc.)
The goal is to be a well-defined entity, not just a website. Knowledge graph thinking forces that clarity.
Structured data and AI citations
Knowledge graphs and structured data are deeply connected. Schema.org markup is how you explicitly tell machines what your content is about, and that signal matters more than ever now that AI models are choosing which sources to cite.
How schema.org markup helps AI models understand your content
When you add JSON-LD structured data to your pages, you are not just helping Google render rich snippets. You are making your content machine-readable in a way that feeds directly into knowledge graph structures. AI retrieval systems, the ones that power Perplexity's citations, Google AI Overviews, and Copilot's source selection, parse structured data to understand what a page is, who wrote it, what entity it belongs to, and how authoritative it is.
Without structured data, an AI system has to infer all of this from unstructured text. That is harder and less reliable. With schema markup, you are handing the AI a cheat sheet.
JSON-LD best practices for AI visibility
JSON-LD (JavaScript Object Notation for Linked Data) is the format Google recommends, and the one AI systems handle best. A few practices that matter specifically for AI visibility:
Keep your JSON-LD in the page head. Retrieval systems parse it early, before they process the full body content. Placing it in the <head> tag ensures it gets read even if the crawler does not render the full page.
Be specific, not generic. A vague Organization schema with just a name and URL does little. Fill in description, foundingDate, sameAs links to your social profiles, and areaServed. The more attributes you define, the richer the entity signal.
Use nesting when it makes sense. A Product schema nested inside an Organization, with review ratings attached, creates a richer graph of relationships than flat, disconnected schemas.
Validate with Google's Rich Results Test. Errors in your structured data mean AI systems may ignore it entirely.
Which schemas matter most for AI visibility
Not all schema types carry equal weight for AI citation. Based on what retrieval-augmented AI systems currently prioritize:
Organization is foundational. It establishes your brand as an entity with attributes, relationships, and provenance. Every business site should have this at minimum.
FAQ is powerful for conversational AI. When someone asks ChatGPT or Perplexity a question that matches your FAQ markup, the structured question-answer format makes it trivially easy for the AI to extract and cite your response.
Product with pricing, availability, and review data helps AI systems recommend specific products. If you sell anything, this schema feeds directly into purchase-related queries.
HowTo structures step-by-step content in a way AI can parse and present. Guides, tutorials, and process documentation benefit significantly from HowTo markup.
LocalBusiness matters for geo-specific queries. When someone asks an AI "best coffee shop in Lyon" or "plumber near me", LocalBusiness schema with proper address, hours, and service area is what helps you surface.
How Google AI Overviews uses structured data
Google AI Overviews (formerly SGE) is particularly interesting because Google built it on top of their existing Knowledge Graph infrastructure. When AI Overviews generates a response, it draws heavily from sources that have clear structured data. Pages with FAQ schema get cited for question-answer queries. Pages with Product schema surface in comparison responses. Pages with HowTo markup appear in procedural answers.
The pattern is clear: Google AI Overviews trusts sources it can understand programmatically. Structured data is how you make your content programmatically understandable. If your pages lack schema markup, you are asking Google's AI to do extra work to figure out what your content is about. Most of the time, it will just pick a competitor who made that job easier.
