> ## Documentation Index
> Fetch the complete documentation index at: https://docs.frictionai.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Entity

> Measure whether AI models understand your brand as a real-world entity before and after live search.

# Entity

Entity measures whether AI models know your brand as a real-world entity. It separates foundation signals, training-data understanding, and live web search uplift.

<Frame>
  <img src="https://mintcdn.com/friction/GwaVVezuSSMLCKiT/images/entity-understanding.svg?fit=max&auto=format&n=GwaVVezuSSMLCKiT&q=85&s=11d8e3e2b8ab7aec7ced2f6df4aaf73e" alt="Entity understanding visual with foundation, training data, and web search layers feeding an entity score" width="960" height="520" data-path="images/entity-understanding.svg" />
</Frame>

<Info>
  Entity understanding is part of brand health and is available on Growth, Professional, and Enterprise plans.
</Info>

## The three layers

<CardGroup cols={3}>
  <Card title="Foundation" icon="database">
    Structured identity signals such as brand name, domain, schema, and known profile data.
  </Card>

  <Card title="Training data" icon="brain">
    What models appear to know before live browsing or retrieval changes the answer.
  </Card>

  <Card title="Web search" icon="search">
    How live sources change or improve the brand answer.
  </Card>
</CardGroup>

## Why it matters

If models do not understand the brand as an entity, prompt and shopping performance can be unstable. Entity gaps often show up as:

* Incorrect or incomplete brand descriptions
* Confusion with similarly named companies
* Weak association with the right category
* High dependence on live search to answer basic brand questions
* Provisional or low-confidence reconciliation

## How to use the page

<Steps>
  <Step title="Read the entity score">
    Start with the headline score and search uplift to see whether live search materially improves the answer.
  </Step>

  <Step title="Inspect layer breakdowns">
    Check foundation, training data, and web search separately. Each layer points to a different fix.
  </Step>

  <Step title="Filter by provider">
    Provider-specific issues can reveal whether a model knows the brand or relies on retrieval.
  </Step>

  <Step title="Use run history">
    Compare historical snapshots to see whether entity understanding is improving after content and structured-data work.
  </Step>

  <Step title="Open prompt diagnostics">
    Use the prompt inspector when a layer needs evidence.
  </Step>
</Steps>

## Fix paths

<AccordionGroup>
  <Accordion title="Foundation is weak">
    Improve structured brand signals: website schema, consistent names, organization data, social profiles, and canonical domains.
  </Accordion>

  <Accordion title="Training data is weak">
    Increase durable public references and clear category association. This may take longer to influence models.
  </Accordion>

  <Accordion title="Web search uplift is high">
    Live search is doing most of the work. That can be good, but it also means source quality and freshness matter heavily.
  </Accordion>

  <Accordion title="Reconciliation is provisional">
    AI may have found a likely match but not enough evidence to strongly confirm it. Check naming ambiguity and entity source consistency.
  </Accordion>
</AccordionGroup>