Interrogating Google’s Knowledge Graph (Manual Practice Example)

The quickest way to understand how Google sees your brand is to ask it directly. This practical lesson shows how to interrogate Google’s Knowledge Graph and its natural language tools to check whether you are recognised as an entity, how confidently, and which topics your pages actually communicate. Using a well-known author as the worked example, it turns abstract ideas about entities into diagnostics you can run on your own site.

What this lesson covers

• Querying the Knowledge Graph Search API and reading its response
• What entity IDs, semantic types and confidence scores reveal
• Extracting entities from page text with the Natural Language API
• Using salience scores to judge a page’s primary topic
• NLP features that shape how language models understand entities
• Common problems in real page analysis and how to fix them

The lesson begins with a raw response from the Knowledge Graph Search API. It walks through the entity name and description, the unique identifier that signals an established entity, the hierarchical type classification, the detailed description with its source, and the confidence score. The score is the key metric: well-established entities score very highly, while new brands tend to score low or are not fully recognised. That makes the API a useful diagnostic for checking your own entity status.

Next, the Natural Language API is used to break a passage of text into entities, each with a type and a salience score between zero and one. High salience marks the primary topic, medium salience marks supporting topics, and low salience marks words with little semantic weight. This matters for SEO because a brand name with low salience on its own page suggests Google does not see it as the main subject. Navigation elements or stray words that outrank core entities point to unfocused content. The recommended check is to run your top pages through the analysis and confirm that primary entities have high salience, a Knowledge Graph ID and the correct type. The lesson then covers features such as position mapping, mention classification and tracking several mentions of the same entity, which shape how models process entities. A real website analysis shows typical issues such as concatenated text, duplicated entities and HTML artefacts, and why cleaning markup and refocusing content helps.

This lesson is part of the first core module, Understanding Entity-Based Search in the LLM Era. It suits SEOs and content teams who want a hands-on method for seeing their pages and brand through Google’s eyes before planning entity improvements.