Thinking in entities, not just keywords

An entity is an identifiable object in the real world — a person, a brand, a place, a product, a concept — that search engines and generative AI recognize and connect to one another, regardless of the exact words used to describe them. Thinking in entities rather than isolated keywords changes how you organize content, and becomes central once that content also needs to be understood by a generative engine.

What an entity looks like in practice

Vurto is an entity (a brand). Paris is an entity (a place). Google Search Console is an entity (a product). Generative engine optimization is an entity (a concept), just like its more common abbreviation, GEO. A keyword, by contrast, is just a string of characters: “GEO” and “generative engine optimization” are two different keywords, but they can point to the same entity if the engine manages to establish the equivalence.

This ability to connect several phrasings to a single object is what distinguishes keyword-based search from entity-based search: Google (through its Knowledge Graph, since 2012) and the language models behind ChatGPT or Claude no longer look only for textual matches — they try to identify what is actually being talked about.

The knowledge graph: connecting entities to one another

A knowledge graph associates each entity with attributes (a brand has an industry, a founding date, products) and with relationships to other entities (a brand belongs to an industry, competes with other brands, is based in a place). When an engine processes a query, it doesn't just look for the words it contains: it tries to map it to one or more entities in its graph, then pull in related entities that can enrich the answer.

This is the mechanism behind why searching for a niche brand often surfaces, alongside the results, its competitors, its industry, or related products: the engine is navigating the graph, not just an index of words. The same logic applies to generative engines: when asked to compare two brands, a model draws on whatever relationships it has been able to establish between the corresponding entities, not solely on text it happens to have seen mentioning both names together.

Take an example: a search for “Claude” can refer to a first name, a Roman emperor, or an AI model built by Anthropic. The engine resolves this ambiguity using context — the other words in the query, browsing history, the relative popularity of each entity — and maps the query to the most likely entity before even looking for pages to display. Content that never explicitly states which entity it's talking about makes this disambiguation harder, at the cost of its own visibility.

Why keywords alone no longer cut it

A content strategy built solely on lists of isolated keywords often produces redundant pages: several similar pages targeting variants of the same word, without real added value between them. A strategy built on entities naturally groups these variants around a single topic covered in depth, which reduces cannibalization and makes it easier for the engine to understand the subject.

It's also a shift in framing: instead of asking “which keyword should I target”, the question becomes “which entities should my content cover, and how do I connect them explicitly”. A concrete example: a site that publishes twelve different pages around variants of the word “GEO” without linking them together dilutes its authority across twelve competing URLs instead of concentrating it on one clearly established entity — a common cannibalization scenario for sites that still think purely in keywords.

What this changes for writing

In practice, thinking in entities means:

• Naming important entities explicitly rather than relying on pronouns or vague paraphrases throughout an article (“that kind of tool” instead of repeating “Google Search Console”).
• Clearly stating the relationships between entities: that a tool belongs to a category, that a method is an alternative to another, that a concept is a subset of a broader one.
• Staying consistent in the terminology used to name a given entity across an entire site, rather than alternating synonyms that blur the association.
• Structuring content so a central entity is covered in depth in one place, with links to related entities — exactly the principle behind the Hub & Spoke architecture.

An example applied to internal linking

In practice, across a content cluster dedicated to SEO, the entity “Google Search Console” should be named consistently on every page that mentions it, with a link pointing to the page that covers it in depth rather than a different definition each time. Conversely, calling the tool “Search Console” in one place, “the Google search console” in another, and “GSC” elsewhere, without ever explicitly establishing the equivalence, makes the matching work harder for an engine that has to decide whether these three mentions really refer to the same thing.

Entities and GEO: an even more direct stake

For a classic search engine, misidentifying an entity costs you a ranking. For a generative engine, it can cost you an entire citation: if the model can't connect your content to the entity it's trying to describe in its answer, it will simply look for that information elsewhere. Structuring an article for GEO and producing GEO-ready content both rely heavily on this entity-by-entity clarity.

The full vocabulary of these notions — entity, knowledge graph, and the other terms of GEO — is gathered in the glossary of GEO and generative AI terms, useful for checking a definition while writing.