Entity Audit

An entity audit is a systematic check of how machines understand the entities on a website: the brand, its products, people, places and concepts. It asks whether each important entity is recognised, clearly defined, consistently named, connected to related entities, and corroborated by sources outside the site. The output is a diagnosis and a prioritised list of fixes.

How it works

MLforSEO’s entity audit material starts with two Google APIs. The Google Cloud Natural Language API extracts the entities in a page and gives each a salience score; the Knowledge Graph Search API checks whether those entities exist in Google’s Knowledge Graph and returns their identifiers. Typical steps include:

• Inventory: build an entity catalogue across six primary categories (organisation, leadership, products and services, locations, events, concepts) and compare it with what the Natural Language API actually finds.
• Salience check: are your key entities the main subject of their pages, or do navigation elements and boilerplate outrank them? See entity salience score.
• Consistency: is every entity named the same way across pages, schema and external profiles?
• Relationships and hubs: does every mention link to a definitive page, and are there orphan entities?
• Structured data: audit schema markup for coverage, connections via @id and sameAs, and errors.
• Corroboration: check Wikidata, Knowledge Panels, directories and press for independent confirmation.

The AI-powered Entity Audit course organises this into five dimensions, the 5Cs of entity representation: clarity, coverage, consistency, connection and corroboration.

How it’s used in SEO and marketing

• Diagnosing why AI systems describe a brand vaguely, wrongly or not at all.
• Competitor comparison of entity coverage and salience.
• Planning entity hubs, schema work and digital PR as one roadmap.
• Re-running on a weekly, monthly and quarterly cadence to track entity health over time.

Example

An audit of a SaaS site finds its proprietary “AI engine” mentioned across dozens of pages but defined on none, author names spelled three different ways, and a salience report where “cookie” and “menu” outrank the product. The fixes are a definitive product hub, one canonical name per person, and template changes that reduce boilerplate in the extracted text.

Limitations and things to watch

• NLP APIs are proxies. They show how one system reads your text, not exactly how Google ranks or how an LLM answers.
• Salience and confidence scores are relative and change between API versions.
• An audit is a snapshot. Without a cadence and an owner, findings go stale.

Learn it on MLforSEO

• How to Audit Entity Structure in Modern Search
• 5 Steps to Identify and Fix Entity Recognition Gaps That Block LLM Visibility
• Entity Audit Cadence: What to Check Weekly, Monthly, Quarterly
• Schema Entity Audit Template
• The AI-powered Entity Audit (course)
• AI Search & LLMs: Entity SEO and Knowledge Graph Strategies for Brands (course)