Entity Salience Score

The entity salience score is a number between 0 and 1 that Google’s Cloud Natural Language API attaches to every entity it extracts. It estimates how important or central that entity is to the whole text you analysed. It is the concrete, measurable version of the broader idea of salience.

How it works

When you send text to the API’s entity analysis method, each entity in the response comes with a name, a type (Person, Organization, Location, Work of Art and so on), its mentions, optional metadata such as a Knowledge Graph ID and Wikipedia URL, and a salience value. Google’s documentation describes scores near 0 as less salient and scores near 1 as highly salient. Within one document the scores are relative: they rank entities against each other, so a score only means something next to the other entities in the same text. Google does not publish the exact formula. In practice, prominence in the text, repeated mentions and the entity’s role in sentences all appear to push it up.

How it’s used in SEO and marketing

• Page focus checks. Does the page’s primary entity get the highest salience, or does something else dominate?
• Weighting in classification. In entity-based search intent classification, more salient entities can carry more weight in the final decision.
• Internal linking. Pages where an entity is highly salient are good link targets for that entity.
• Before and after comparisons when you rewrite a page to make its subject clearer.

Example

MLforSEO’s knowledge graph lessons analyse a short paragraph about Jane Austen. The API returns Jane Austen as a Person with the highest salience, Pride and Prejudice as a Work with medium salience, and generic terms such as “world” with low salience. A common audit finding works the other way: run a SaaS homepage through the API and repeated navigation labels can outscore the brand name, which tells you the boilerplate is diluting the page’s main subject.

Limitations and things to watch

• Short texts such as single keywords produce little variation. For large keyword datasets, MLforSEO recommends a prevalence measure (how many keywords mention an entity) alongside or instead of raw salience.
• Scores depend on exactly what text you send. Strip navigation and footers, or analyse the main content only, if you want to judge the article itself.
• Salience is Google Cloud’s model output. It is a useful proxy for how machines read a page, not a ranking factor Google has confirmed for Search.
• Thresholds such as “primary entity above 0.8” are working heuristics. Compare pages of similar length and structure rather than chasing a fixed number.

Learn it on MLforSEO

• Why Navigation Elements Outrank Your Brand in Entity Analysis
• Entity-Based Search in the LLM Era: How Google Structures and Interprets Entities
• 5 Steps to Identify and Fix Entity Recognition Gaps That Block LLM Visibility
• Extracting Entities with Google Cloud Natural Language API in Python (Tutorial)
• The Semantic AI-powered SEO Keyword Research course, which uses salience in query entity analysis