Skip to main content

Insight · September 14, 2026

Become a thing,
not a string.

AI search does not cite pages it cannot place. It cites entities it can verify. Entity SEO is the work of becoming one the machine is sure of.

01 · The shift

Google stopped matching words. It started recognizing things.

On May 16, 2012, Google announced the Knowledge Graph under a phrase that still explains the whole discipline: things, not strings. Amit Singhal, then head of search, described a move away from matching the letters you typed toward recognizing the thing you meant. It launched with 500 million entities and 3.5 billion facts about how they connect.

That graph did not stay small. By May 2024 it held more than 1.6 trillion facts about 54 billion entities, up from 500 billion facts on 5 billion entities in 2020. The search box quietly stopped being a keyword matcher and became a lookup against a map of the world. Entity SEO is the work of making sure your brand is on that map.

02 · What it is

Entity SEO makes your brand a thing the machine is sure of.

In one of Google's own patents, an entity is defined as a thing or concept that is singular, unique, well defined, and distinguishable. A person, a company, a product, a place. A keyword is a string of characters. An entity is a node: it has an identity, a set of facts, and links to other nodes.

Entity SEO is the work of getting your brand, its people, and its products recognized as distinct entities in that graph, with facts that agree everywhere a machine looks. It is not about repeating a phrase until you rank for it. It is about being an unambiguous thing the search engine, and now the answer engine, can point to with confidence.

The order of operations changed. Google increasingly resolves the entity a query is about first, then surfaces the pages most associated with that entity. So your representation in the graph is not a nice to have on top of ranking. It is the thing that decides whether your pages are eligible to be surfaced at all, and whether an answer engine treats you as a source it can name.

The one line to keep

“A keyword is something you rank for. An entity is something the machine is sure exists.”

03 · What builds an entity

An entity is built from signals that agree.

01

A stable identity

One name, one canonical page, one identifier. Google assigns a KGMID, Wikidata assigns a Q number. The machine needs a single place to hang every fact about you.

02

Structured facts

Schema markup that states what you are, who runs you, and what you make, in a form a machine reads without guessing.

03

Corroboration

The same facts repeated on sources Google already trusts. Wikipedia and Wikidata feed the graph directly, so a mention there counts for more than another claim from you.

04

Connections

sameAs links that tie every profile of you back to one node, so the machine does not split you into several half known things.

05

Consistency

The name, the category, and the facts match everywhere. A contradiction reads as two weak entities instead of one strong one.

04 · Why it matters now

AI answers cite entities, and the graph just got stricter.

In June 2025, Google pruned the Knowledge Graph, removing more than three billion entities in a single week. Across two closely timed updates the graph contracted by about 6.26 percent, and the count of generic entities merely labeled a thing fell by more than 15 percent. This was not a loss of data. It was a move for clarity: a leaner, higher confidence graph to power AI Overviews, AI Mode, and Gemini. Much of what went was the vague and the temporary, including pandemic era event entities that no longer earned their place. Google kept the entities it could stand behind.

That matters because a model cannot verify an unfamiliar brand. As these systems lean on a slimmer graph, only the best labeled, most authoritative entities get surfaced as sources. If your brand is not clearly established as a recognized entity, with consistent signals across your own site, your structured data, and the sources Google trusts, you risk being absent from the answer, not ranked below it.

05 · One example that shows the difference

Two brands, one query, one citation.

A buyer asks an assistant to name the best options in your category. The model does not read the whole web in that moment. It answers from what it can place. A brand that exists as a clear entity, with a category, a set of facts, and links a machine can follow, is a candidate it can name and cite. A brand that exists only as pages of keywords is a string the model cannot vouch for, so it gets left out of the answer it could have led. Same product. One is citable. The other is invisible.

The fix starts with telling the machine plainly what you are. The minimum is a piece of structured data that states your identity and points every other profile of you back to the same node.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Your Brand",
  "url": "https://yourbrand.com",
  "sameAs": [
    "https://www.wikidata.org/wiki/Q...",
    "https://www.linkedin.com/company/yourbrand",
    "https://www.crunchbase.com/organization/yourbrand"
  ]
}

One node · every profile pointing back to it

06 · The work

You do not write your way in. You get confirmed in.

The work is not more words on more pages. It is agreement. Pick the canonical page for each entity you own: the brand, the founders, the products. State the facts once, in schema and in plain language, and in the same form everywhere else a machine will check.

Then reach outward. Claim the profiles that feed the graph and point them all back with sameAs. If your entity qualifies, Wikidata and Wikipedia feed Google directly, so a correct entry there compounds. Earn mentions on sources Google already trusts, because corroboration from outside you counts for more than another claim from you.

Keep it consistent. The fastest way to weaken an entity is to describe it two different ways, because the machine then has to decide whether you are one thing or two, and it hedges by trusting you less.

07 · The honest part

Entity strength is earned, not declared.

Schema markup does not mint an entity. It describes one the machine still has to believe. You cannot mark up your way into the graph if nothing outside your own site corroborates you, and you cannot force a knowledge panel into existence. The signals have to be real: a genuine presence, cited by others, consistent over time.

So entity SEO is slower than chasing a keyword, and it compounds in a way keywords never did. A keyword ranking is settled again every time the query runs. An entity, once the machine is sure it exists, becomes a standing fact it stops questioning. You are not renting a position. You are becoming part of the map the answer is drawn from.

Closing

A model cannot cite what it cannot place.

Search your own brand in an answer engine and read what it says back. If it hedges, or names a competitor instead, you are a string it could not place. The work from there is not more content. It is making yourself one clear thing, everywhere the machine looks.

Knowledge Graph launch and scale · Google and Search Engine Land · June 2025 clarity cleanup figures · Search Engine Land and SOCi · entity definition from Google's question answering patent · entity signals via Schema App and Similarweb · reviewed September 2026

Share this perspective

More insights

Adjacent perspectives.

How to Choose an AI Agent Framework

8 min read

How to Choose an AI Agent Framework

A year ago the framework you picked barely mattered. Every demo was one agent calling a few tools, and you could rewrite it in an afternoon. That is over. As teams move from demos to production, the framework decides how the agent holds state, where a person can step in, and what you can see when it fails, and rewriting it becomes a quarter, not an afternoon. Five names carry most of the weight right now: LangGraph, a low level framework for long running stateful agents; CrewAI, a standalone framework for role based crews; the OpenAI Agents SDK, released March 2025 as the successor to the experimental Swarm and provider agnostic across 100 plus models; Google ADK, a code first toolkit for teams inside Google Cloud; and the Microsoft Agent Framework, which folded AutoGen and Semantic Kernel into one path at its October 1, 2025 preview and sent both predecessors to maintenance mode. Which is best is the wrong question. The four traits that separate a demo from a system, and how to pick the one you can leave.

What Is Context Rot

8 min read

What Is Context Rot

A model with a million token window still does not read it evenly. It handles the front and the back with care and grows careless in the middle, and the more you load in, the wider that careless zone gets. That is context rot, the measurable drop in quality as the input grows, and it starts long before the window is full. In a July 2025 report, Chroma tested 18 current models, among them GPT 4.1, Claude 4, Gemini 2.5, and Qwen3, and found every one got worse as the input got longer, even on finding a fact or copying text. Two years earlier a Stanford team named the lost in the middle effect, where accuracy fell more than 30 percent when the answer sat in the middle rather than the edges. What context rot is, why a bigger window is not a bigger memory, and the smaller cleaner window that fixes it.

Bttr. Field Brief

The brief Bttr. writes for senior buyers.

Monthly. One signal worth your time on Brand Operating Systems, AI search visibility, and the infrastructure buildout. No filler.

Industries We Serve

Aerospace & DefenseBiotechnologyMedical & HealthcareManufacturingFinancial ServicesConsumer ProductsEnterprise Software

New Business

Start a project

Headquarters

North America

© 2026 Bttr. All rights reserved.