What is a knowledge graph?
Most business data sits in tables. A knowledge graph stores the same kind of facts as a web of connections instead: this product belongs to this brand, is sold in these markets, competes with that product, and appeals to this audience. Each fact is a link between two entities.
The structure that defines which kinds of entities and relationships are allowed is called an ontology. Search engines run large public knowledge graphs, and AI systems use graphs to look up facts and follow connections a plain document search would miss.
How does a knowledge graph work?
| Part | What it is | Example |
|---|---|---|
| Entity | A thing with an identity | A brand, a product, a store, a campaign |
| Relationship | A typed link between two entities | Made by, sold in, competes with |
| Property | A fact about one entity | Launch date, price, category |
| Ontology | The rules for which entities and links exist | A product must belong to exactly one brand |
Questions are answered by following links: which products in this category compete with ours in Canada? Some AI systems combine graph lookups with document retrieval, which is sometimes called GraphRAG.
What is an example of a knowledge graph?
Why an AI assistant describes a brand the way it does
Ask an AI assistant about a mid-size outdoor brand, and its answer pulls from what it has learned and from entity data, such as the brand's category, founding year, parent company, and best-known products.
If that entity data is wrong or thin, for example listing a discontinued product line, the assistant can repeat the mistake. Brands that keep their public entity information accurate give AI systems better facts to draw on.
Knowledge graph vs. semantic layer
| Semantic layer | Knowledge graph | |
|---|---|---|
| Defines | How metrics are calculated | What things are and how they relate |
| Example | ROAS = revenue / ad spend, net of returns | Product X is made by Brand Y and sold in Canada |
| Main user | Dashboards, analysts, and AI asking for numbers | Search, recommendations, and AI asking about entities |
| Together | Gives AI the right math | Gives AI the right facts and connections |
Why do knowledge graphs matter?
They matter in two places for marketers. Inside the business, they connect data that lives in different systems around the entities it describes. Outside it, public knowledge graphs and entity data feed how search engines and AI assistants describe and recommend brands.
- Better AI answers. Following relationships lets AI answer questions that span products, markets, and competitors.
- Brand accuracy in AI search. Accurate public entity data gives assistants better facts about the brand.
- Maintenance is the cost. Entities and links go stale like any data, so someone has to own them.
How does Alli do this?
Alli by PMG's Brand Equity tracks how AI models describe your brand through Share of LLM Presence: whether the brand appears, where it ranks, and how it's described across a range of AI models, against your competitive set and over time.
Inside the business, Alli Semantic Layer models how performance data connects to campaigns, ads, accounts, and calendars, and maps platform naming into one plain-English taxonomy, so every tool reads the same structure.
Explore Alli Brand EquityKnowledge graph FAQ
What is a knowledge graph in simple terms?
It's a map of facts stored as connections: things like brands, products, and places, linked by relationships such as made by, sold in, or competes with.
What is the difference between a knowledge graph and a semantic layer?
A semantic layer defines how business metrics are calculated. A knowledge graph defines entities and the relationships between them. AI systems benefit from both.
What is the Google Knowledge Graph?
Google's large database of entities and facts, which it uses to understand searches and show information panels. Brands appear in it as entities with their own facts.
What is GraphRAG?
An approach that combines knowledge graph lookups with document retrieval, so an AI system can follow relationships between entities as well as find relevant passages.
Last updated Back to the glossary