A knowledge graph is a database that stores facts as connections between things: a company, a person, a place or a product (nodes), linked by named relationships such as “founded by” or “located in” (edges). Google’s Knowledge Graph, Wikidata and the JSON-LD on your own website are all knowledge graphs of different sizes. Search engines and AI systems use them to know which real-world entity a page is about.
Nodes, edges and triples
The basic unit is a triple: subject, predicate, object.
| Subject | Predicate | Object |
|---|---|---|
| Brilad | founded by | Gunakesh Parmar |
| Brilad | founded by | Rohit Bhisey |
| Brilad | located in | Pune |
| Gunakesh Parmar | profile | LinkedIn URL |
Each subject and object is a node; each predicate is an edge. Because nodes are shared, facts join up: once the system knows who founded Brilad and where Brilad is, it can answer questions that combine both.
The knowledge graphs that matter
Google’s Knowledge Graph
Google introduced its Knowledge Graph in May 2012 under the phrase “things, not strings”, meaning a shift from matching words to understanding entities. It powers knowledge panels, the boxes of facts that appear beside some search results. Entities in it have machine IDs that can be looked up through Google’s Knowledge Graph Search API.
Wikidata
Wikidata, launched by the Wikimedia Foundation in 2012, is an open knowledge graph that anyone can edit, with each entity identified by a Q number. Its data is reused widely, including by search engines, and it appears in the text collections language models learn from.
Your own graph
The JSON-LD on your website is a small knowledge graph about your business. When its nodes use stable @id values and link out through sameAs to the same entities elsewhere, it connects to the larger graphs. The JSON-LD guide shows how.
Knowledge graphs and AI answers
Language models store knowledge in their weights, which makes it hard to check and update. Knowledge graphs store explicit facts that can be checked and changed. Many AI systems combine the two: they use a graph or structured sources to identify entities and confirm facts, and the language model to write the answer. A company that is clearly represented in these sources is easier to describe correctly.
How to get your business represented
- Publish the basic facts clearly on your own site: an About page, a contact page, founder profiles.
- Add Organization and Person schema with sameAs links.
- Create and verify a Google Business Profile.
- Keep profiles on LinkedIn, Crunchbase and industry directories consistent.
- Earn independent coverage, which is what makes a Wikidata item and a broader knowledge panel possible.
- If Google shows a knowledge panel for you, claim it through Google’s verification process so you can suggest corrections.
Frequently asked questions
Can I pay to get a knowledge panel?
No. Google creates panels from its own data. Services that promise panels for a fee cannot guarantee them.
How do I know if Google sees my company as an entity?
Search your exact company name and look for a panel, or query the Knowledge Graph Search API. A local panel from Google Business Profile is a common first sign.