Entity Graph

Definition #

Definition: An entity graph is a structured representation of named entities and the typed relationships between them. Nodes represent real-world things — organizations, people, products, locations, or concepts — while edges represent relationships such as a company employing a person, offering a product, belonging to a category, or being identified by a public reference.

Definition: The entity graph separates the entity from the documents that mention it. A page is a document. An entity is the person, organization, product, or concept represented independently of any single document.

Definition: Entity graph is an established industry term. Machine Relations uses it both for graph-based representations in named knowledge and reconciliation systems and for a brand-side diagnostic map of public identity relationships. This usage does not claim that every search engine or answer engine implements the same graph or uses it at the same stage.

Named Product Boundaries #

Premise: The Google Knowledge Graph Search API finds matching entities in Google’s Knowledge Graph. Google documents real-world people, places, and things as graph nodes, and the API uses schema.org types and JSON-LD. An API result shows that a matching Knowledge Graph entity was returned; it does not by itself establish public Google Search query routing, identity correctness for every query, or treatment by an answer engine (Google Knowledge Graph Search API).

Premise: Google Cloud Enterprise Knowledge Graph documents entity reconciliation for enterprise data. It maps records from BigQuery tables and clusters candidate matches using fuzzy text, relationships, entity types, and attributes. This product scope does not establish how public Google Search or any answer engine retrieves, selects, cites, or recommends web sources (Google Cloud Enterprise Knowledge Graph).

Premise: Google Search Central describes structured data as a standardized format that gives explicit clues about page meaning. Organization markup can help Google understand administrative details and disambiguate an organization in Search, but Google does not guarantee that structured-data features will appear and recommends measuring effects before and after a change (Google Search Central: Organization markup).

Premise: Schema App argues that schema markup and content knowledge graphs can help AI systems interpret entities and relationships. That article is an attributed vendor and practitioner perspective, supported partly by the company’s hypotheses and customer examples; it is not independent proof of universal large-language-model retrieval, source-selection, authority, or confidence mechanisms (Schema App, vendor perspective).

How Machine Relations Uses the Model #

Definition: Entity resolution is the task of deciding whether records or mentions refer to the same entity.

Premise: Google Cloud documents this task for enterprise record reconciliation. That product example should not be expanded into a claim that public search replaces keyword matching with entity-node routing.

Premise: Machine Relations treats answer-engine use of public entity relationships as a hypothesis to test by engine and query. The observable questions are whether the intended entity is resolved, which documents are retrieved, which hosts and URLs are cited, and whether the answer attributes or recommends the entity. The graph alone does not reveal an engine’s source-selection or authority logic.

Premise: Consistent identifiers and relationships can be audited as candidate explanations for correct identity resolution. Relationship count is not a published cross-engine “resolution confidence” score, and a visibility outcome does not identify the graph as its cause.

Brand-Side Entity Graph Elements #

Premise: Google Search Central recommends Organization structured data to help Google understand an organization’s administrative details and disambiguate it in Search (Google Search Central: Organization markup). A brand-side entity record can document:

• Organization identity. Canonical name, aliases, official website, founding date, and headquarters location.
• People and roles. Founders, executives, and subject-matter experts with current titles and identifiers.
• Products and services. Named offerings, descriptions, and the organization that provides them.
• Category associations. Industry classifications, service categories, and intended market positioning.
• Corporate relationships. Parent companies, subsidiaries, partners, and acquired entities.
• External identifiers. LinkedIn profiles, Crunchbase entries, Wikidata QIDs, and similar references, with provenance, ownership, and verification status recorded separately.
• Independent corroboration. Sources with editorial or institutional independence that support a specific identity or relationship claim.
• Observed answer-engine behavior. Declared query runs showing resolution, retrieval, citation, attribution, or recommendation without treating those observations as proof of mechanism.

The record should keep four evidence roles distinct: owned declarations, external identifiers, independent corroboration, and observed answer-engine behavior. A directory or profile can identify an entity without independently verifying every claim on the profile. A brand’s preferred category description is not automatically an externally established classification.

Entity Graph and Entity Clarity #

Definition: Entity Graph names the relationship structure. Entity Clarity names the coherence of the public identity represented through that structure.

Premise: Within the Machine Relations model, entity-graph work belongs to the Entity Clarity layer. The graph is a diagnostic map for finding conflicting names, missing relationships, ambiguous identifiers, unsupported category claims, and evidence-role confusion across owned and third-party surfaces.

Working hypothesis (not measured): Compatible identifiers and relationships across public sources may help a particular system resolve the intended entity. Test that hypothesis directly by engine and query rather than treating consistency, structured data, public profiles, or earned coverage as a universal prerequisite for retrieval or citation.

Audit Method #

Start with a canonical entity record that lists the organization’s name, key people, products, category, and identifiers. Then compare that record against:

1. Owned declarations. Visible website content, JSON-LD, schema.org types, and sameAs references.
2. External identifiers. LinkedIn, Crunchbase, Wikidata, and industry-directory records, each labeled by provenance and ownership.
3. Independent corroboration. Media, institutional, regulatory, or other independent sources that support specific facts.
4. Observed answer-engine behavior. Repeatable query runs that record entity resolution, retrieved hosts, cited URLs, attribution, and recommendation separately.
5. Product pages and service descriptions. Internal consistency of names, categories, and relationship claims.

Mark conflicts, stale relationships, unsupported category claims, missing identifiers, and evidence-role mismatches. Preserve citations for externally asserted facts. The audit output is a gap list, not a confidence or authority score. Each gap names a relationship, identifier, or observation that is missing, inconsistent, or not yet independently supported.

Practical Boundary #

Definition: An entity graph is a representation, not proof of visibility or recommendation. Adding a relationship to owned content or structured data does not establish that an external system accepted it.

Premise: A graph is intended to make relationships explicit. Whether it changes resolution, retrieval, citation, attribution, or recommendation requires direct observation.

Premise: Schema markup is one standardized machine-readable way to declare entities and relationships on a page. It does not replace visible-content consistency, establish itself as the primary mechanism used by every system, or guarantee a Search feature, answer-engine citation, or recommendation.

Premise: Independent corroboration, owned declarations, public identifiers, and earned coverage can each play different evidence roles. No source cited here establishes earned media as a universal prerequisite for entity recognition, retrieval, citation, or recommendation.

Frequently Asked Questions #

What is the difference between an entity graph and a knowledge graph? #

Machine Relations definition: A knowledge graph is the broader graph of entities, relationships, and facts maintained for a system or dataset. An entity graph is the relationship view centered on a particular entity — for example, an organization, its people, products, category, and identifiers. This distinction is a Machine Relations convention for this page.

How does an entity graph affect AI search visibility? #

Premise: The graph supplies a diagnostic model, not a universal visibility mechanism. Entity-resolution systems can use relationships and attributes to match records. For an answer engine, measure whether the intended identity is resolved and which sources are retrieved or cited; do not infer source authority, selection logic, or a confidence score from the graph alone.

Can a brand build its own entity graph? #

Premise: A brand can maintain owned declarations, structured data, visible content, identifiers, and a record of third-party references. It cannot declare that an external system accepted those relationships. Owned declarations, external identifiers, independent corroboration, and earned coverage should be recorded as separate evidence roles; none is treated here as a universal prerequisite.

What is the relationship between entity graphs and schema markup? #

Definition: Schema markup, commonly encoded as JSON-LD with schema.org vocabulary, is one machine-readable way to declare entities and relationships on a page. Organization, Person, Product, and other types can express parts of the conceptual graph.

Premise: Google says structured data provides explicit clues and may enable Search features, but appearance and external acceptance are not guaranteed (Google Search Central: Organization markup).

How do you measure whether an entity graph is working? #

Premise: Measure observable states separately. The Google Knowledge Graph Search API can show whether a matching entity is returned from Google’s Knowledge Graph; that result does not confirm Search treatment or answer-engine behavior (Google Knowledge Graph Search API). For answer engines, record identity resolution, retrieved host, exact cited URL, attribution, and recommendation under a declared query protocol. Treat movement after a graph change as observed correlation unless a controlled intervention isolates that change from other candidate causes.