Stop Optimizing for Keywords. Start Being an Entity.

For twenty years, the mental model of getting found was keywords: pick a phrase, build a page around it, repeat it enough, climb the rankings. That model is now actively counterproductive in AI search. The Princeton/Georgia Tech GEO study found that keyword stuffing measurably reduced a page's visibility in generative engines — the one tactic that clearly backfired. The replacement model is worth internalising, because everything else follows from it: engines recommend entities, not keywords.

Keyword thinking vs entity thinking

A keyword is a string of text. An entity is a thing the engine knows about — a specific business with a name, a location, a category, a reputation, and relationships to other known things. Google has spent a decade building a Knowledge Graph of entities; AI engines reason over the same kind of structure. The question has changed from "does this page contain the phrase 'plumber Calgary'?" to "is there a real, identifiable plumbing business in Calgary I'm confident enough about to recommend by name?"

The difference in practice:

• Keyword thinking: "I need a page targeting 'emergency plumber Calgary.'"
• Entity thinking: "I need to be a recognised plumbing business in Calgary — verifiable, consistent, and connected — so that when anyone asks, I'm a name the engine can safely give."

The first optimises a page. The second optimises you. Only one of them survives a system that writes answers instead of listing links.

What makes you an entity

Becoming recognisable as an entity is concrete work, not vibes. It's the Recognize layer of the REAL Method, and it has a handful of components:

• A consistent identity (NAP). Your name, address, and phone must be identical across your site, Google Business Profile, and every directory. Inconsistency forces the engine to wonder whether "Westside Plumbing Co." and "Westside Plumbers" are one business or two — and uncertainty is what keeps you out of answers.
• Structured self-description. A LocalBusiness subtype in your schema (Plumber, not generic Organization) and a clear About page state, in machine-readable terms, what you are and where you operate.
• sameAs links — the connective tissue. This property points from your site to your other authoritative profiles: Google Business Profile, social accounts, and, ideally, a Wikidata entry. sameAs is how you tell an engine "all of these describe the same real entity." Wikidata matters in particular because it's a foundational data source for Google's Knowledge Graph — a Wikidata Q-ID is a machine-readable anchor for your existence.
• Corroboration. Reviews, mentions, and citations across independent sources that all point at the same identity. Recognition and evidence reinforce each other: the more places that describe you consistently, the more confidently the engine can place you.

The disambiguation problem nobody thinks about

Here's the failure mode entity thinking guards against. Type a generic business name into an engine and watch it hedge — because there might be five businesses with similar names across the country. If the engine can't tell which one you are, it won't risk naming you in a local answer. Strong entity signals are what let it resolve the ambiguity in your favour.

This is why two businesses with identical services and similar reviews can have completely different AI visibility. The one with consistent NAP, a specific schema type, and rich sameAs links is legible. The other is a blur the engine declines to guess at.

How to make the shift, practically

You don't need to rebuild your site. You need to make yourself unmistakable:

1. Audit your NAP across your site, Google Business Profile, and your top directories. Fix every inconsistency, down to "Street" vs "St."
2. Set a specific schema type and add sameAs links to all your real profiles. The Schema Generator and GEO Grader handle this.
3. Claim and complete the profiles that anchor your identity — Google Business Profile first, then the platforms that matter in your industry. Consider a Wikidata entry if you have any third-party coverage to cite.
4. Stop writing keyword-stuffed pages. Replace repetition with specifics — real prices, timelines, service areas. Specifics are what get quoted; repetition is what gets penalised.
5. Keep it consistent over time. Entity recognition compounds; every new consistent signal strengthens the others.

The reframe that makes the rest easy

Once you think entity-first, the rest of AI visibility stops feeling like a grab-bag of tactics. Schema, NAP consistency, sameAs, reviews, quotable answers — they're not a checklist, they're all answering one question the engine is asking: can I identify this business clearly enough to recommend it by name? Get that yes, and citations follow. Chase keywords instead, and you're optimising for a system that no longer exists.

See how recognisable you are right now with the free AI Visibility check — Recognize is the first of the four scores it returns. hello@rankinglocal.ai reaches me directly.