Does Google Actually Know Your Business Exists? A Guide to Entity Signals

Two plumbers, same city, same size, same quality of work. Comparable websites, comparable reviews, comparable prices.

Ask an AI assistant who to call and it names one of them, consistently, and never mentions the other.

Two identical plumbers

The instinct is to assume the invisible one has a worse website, or fewer reviews, or has not "done SEO." Frequently none of that is true.

What is usually true is duller. The invisible plumber moved offices in 2023 and forty directories still list the old address. His Google Business Profile says "Bob's Plumbing LLC" while his website says "Bob's Plumbing & Heating" and his invoices say "Bob Plumbing Co." His phone number appears in three variants because a call-tracking number was rolled out to some listings and not others.

None of that is a mistake anyone would notice. All of it is a contradiction to a machine trying to work out whether these are one business or several — and if they are one, which facts are true.

This article is about how that verification works and what you can do about it. It is unglamorous, mostly free, and it is the foundation underneath everything we have written about getting shortlisted by AI agents.

What an entity actually is

Google's own definition is admirably plain: an entity is "a thing or concept that is singular, unique, well-defined and distinguishable."

Your business is an entity. The words on your website are strings — text about that entity. The difference is not academic, because search systems store the two very differently.

Look at the right-hand column. Every one of those phrases appears on thousands of business websites, and none of them identifies anything. A machine cannot verify "trusted local expert" and cannot distinguish you from anyone else claiming it.

The left column is different. Each item is a specific thing that can be checked against other sources, connected to related things, and stated with confidence.

The strategic implication: writing more marketing copy adds strings. Making your facts specific, consistent and machine-readable strengthens the entity. Only one of those changes whether you get recommended.

Where this came from, briefly

This is not new, which is worth knowing because it means the mechanics are stable rather than a passing trend.

On 16 May 2012, Google published a post by Amit Singhal titled "Introducing the Knowledge Graph: things, not strings." At launch it contained over 500 million objects and 3.5 billion facts about the relationships between them.

Its foundation was Freebase — an open, collaboratively edited database launched by a company called Metaweb in 2007, describing itself as "an open, shared database of the world's knowledge." Google acquired Metaweb in 2010 and built the Knowledge Graph on that groundwork.

The visible surface most people recognise is the Knowledge Panel, the box that appears beside search results for a recognised entity. But the panel is a symptom, not the system. The graph itself runs underneath everything, including the AI features layered on top of search today.

Why this matters more now than it did in 2015

For years entity work was a nice-to-have. A strong entity helped local ranking a bit and occasionally produced a Knowledge Panel. Sites with sloppy data still ranked fine on the strength of their content and links.

Three things changed that.

First, answers replaced links. When a system generates an answer rather than a list, it has to assert things. Asserting requires confidence, and confidence requires sources that agree. A page can rank at position four with ambiguous data; an answer cannot cite you if it is unsure who you are.

Second, the surfaces multiplied. Google's AI features, ChatGPT, Perplexity, Gemini and Copilot all assemble answers from crawlable sources and structured data. Each one independently makes the same judgement about whether your identity is coherent.

Third, and most consequentially, verification became the bottleneck. These systems face real reputational cost for being wrong. Recommending a business with the wrong phone number is a visible failure. So the safe move, when your data conflicts, is simply not to name you.

That silence is what makes entity problems so persistent. A ranking drop shows up in Search Console. Being left out of an answer shows up nowhere at all.

The five layers

Entity strength is built in a specific order, and the order is not arbitrary. Each layer only works if the one above it is solid — schema markup declaring facts that contradict your directory listings actively makes things worse.

Layer 1 — Consistency

This is where most of the damage lives, and it is entirely fixable with patience rather than expertise.

The rule sounds trivial: your name, address, phone number, hours and service area should be identical everywhere they appear. In practice almost nobody achieves it, because business details drift over years without anyone deciding they should.

What counts as a contradiction

More than you would expect. A parser comparing two strings has no judgement, so all of these register as differences:

• "Suite 660" vs "Ste 660" vs "#660" — three different strings describing one place.
• "Bob's Plumbing LLC" vs "Bob's Plumbing" — the legal entity name and the trading name are not interchangeable to a machine.
• Call tracking numbers deployed to some listings and not others. This one is self-inflicted and extremely common.
• An old address still live on directories after a move. The single most destructive item on this list.
• Hours that differ between your site and your Google Business Profile, usually because one was updated and the other forgotten.

The fix

Start by writing down the canonical version of your facts. Once, in a document, including punctuation and abbreviation choices. Then correct sources in order of authority: your own website first, Google Business Profile second, major directories third, the long tail last.

Authority weighting matters here. One contradiction on a major platform does more damage than several on obscure ones, so resist the urge to work alphabetically.

Layer 2 — Structured data

Consistency stops a machine being confused. Structured data stops it having to guess at all.

Schema markup is a small block of JSON in your page source that states your facts explicitly. Your visitors never see it. Every machine reading your site does.

Without it, a system has to infer that the phone number in your footer belongs to the business named in your header at the address in your contact page. Usually it gets there. Sometimes it does not — and "usually" is a poor foundation for something you cannot measure.

{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "Your Business Name",
  "telephone": "+13035551234",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "1001 Bannock St #660",
    "addressLocality": "Denver",
    "addressRegion": "CO",
    "postalCode": "80204"
  },
  "openingHoursSpecification": [{
    "@type": "OpeningHoursSpecification",
    "dayOfWeek": ["Monday","Tuesday","Wednesday","Thursday","Friday"],
    "opens": "08:00", "closes": "18:00"
  }],
  // the line most businesses omit — see Layer 3
  "sameAs": ["https://your-google-profile-url"]
}

The five types with the most impact on entity recognition are Organization or LocalBusiness (your primary entity), Person (ties a named human to the business, which also supports E-E-A-T), FAQPage, HowTo and Product where relevant.

One warning that matters more than it used to: schema must describe what is actually on the page. Markup claiming reviews or content a visitor cannot see is a guideline violation, not a clever shortcut. We audit our own pages for this specifically, because it is easy to let markup and content drift apart during a redesign.

Layer 3 — sameAs, the most underused two lines in SEO

Here is a small thing with disproportionate value.

The sameAs property lets you declare that the business described on your site is the same entity as the one at other verified URLs. Your Google Business Profile. Your LinkedIn company page. Your Facebook page. An industry association listing. A Wikidata item, if one happens to exist.

"sameAs": [
  "https://www.google.com/maps/place/your-listing",
  "https://www.linkedin.com/company/your-company",
  "https://www.facebook.com/YourPage",
  "https://www.bbb.org/us/your-profile"
]

Without it, a machine has to work out on its own that four similar-looking listings describe one business. With it, you have told it. That is the entire mechanism, and it takes about five minutes.

Layer 4 — Third-party corroboration

Everything so far is you describing yourself. Layer four is other people describing you, and machines weight it far more heavily for an obvious reason: self-description is free and unfalsifiable.

Anyone can write "the most trusted plumber in Denver" on their own homepage. Nobody can fake forty independent, timestamped, specific reviews attached to a verified profile.

What counts as corroboration, roughly by strength

• Reviews on verified platforms. The strongest signal available to most local businesses — independent, dated, specific and hard to manufacture at scale.
• Consistent directory listings. Weaker individually, but they compound. Twenty listings agreeing on your address is a strong statement about your address.
• Local press and community mentions. Sponsorships, chambers, associations, local news. Genuinely valuable and routinely ignored.
• Supplier and partner pages. A manufacturer's "authorised installer" list is third-party verification with real weight.
• Professional licensing records. Public, authoritative and permanent. Make sure your listed details match your canonical facts.

Notice how much of that list has nothing to do with marketing. Entity strength is often just the sediment left behind by operating properly for a few years — provided the details agree with each other.

Review recency matters as much as volume, incidentally. Forty reviews from 2022 say something about a business that existed in 2022. Six from the last quarter say something about the business today. The mechanics of building that habit are in our guide to review collection, and the FTC rules covered there apply here too.

Layer 5 — Specificity

The final layer is the one most within your control and most commonly wasted.

Vague claims cannot be cited, because they cannot be verified against a question.

Read the left column aloud and notice that it is how almost all business copy is written. It sounds professional. It is completely inert — every competitor makes identical claims, so none of them distinguishes anyone.

The right column is not better writing. It is better data. And the useful part is that it also converts better with humans, because specificity reads as confidence while vagueness reads as hedging.

Score your entity in three minutes

Fifteen checks, weighted by how much each one actually matters. Tick what is genuinely true — nobody sees the result but you.

Four situations that break the normal rules

The five layers assume a business with one location, one name and a stable history. Plenty of businesses do not fit that, and the standard advice actively misleads them. Here are the four we get asked about most.

The service-area business with no public address

Plumbers, electricians, mobile groomers, consultants working from home. The instinct is to hide the address, and that instinct is correct — but hiding it badly is worse than either alternative.

Configure your Google Business Profile as a service-area business properly, which lets you specify the areas you cover while keeping the street address private. What you must not do is invent an address, use a virtual office you do not staff, or list a co-working space you visit occasionally. Those get profiles suspended, and a suspension is a far bigger entity problem than a hidden address ever was.

Then compensate on layer five. A business without a verifiable address has to be unusually specific about everything else: named service areas, named services, real hours, real response times. Specificity substitutes for proximity.

The multi-location business

Each location is its own entity, related to a parent entity. That relationship needs stating rather than assuming.

In practice: one page per location with genuinely distinct content, one Google Business Profile per location, and schema on each location page that links back to the parent organisation. The failure mode is near-identical location pages differing only by city name — which reads as thin duplication to a search engine and provides nothing specific for an AI to cite.

The bigger hazard is that multi-location businesses accumulate contradictions faster than anyone can fix them. A franchise with thirty locations has thirty opportunities for an old phone number to survive. This is precisely why a well-run small business frequently out-verifies a national chain.

The business that rebranded

Name changes are the hardest entity problem there is, because the old name genuinely existed and the internet has an excellent memory.

Do not attempt to erase the old name — you will fail, and the fragments left behind will contradict you. State the relationship explicitly instead. Keep an "formerly known as" line on your about page. Update the Google Business Profile name rather than creating a new listing. Redirect the old domain rather than abandoning it, because those inbound links are corroboration you already earned.

Expect this to take two to three quarters rather than weeks. The old entity has to decay while the new one accumulates, and there is no way to hurry it.

The business with a common name

If you are "Denver Dental" or "A1 Plumbing," you are competing for identity with businesses that are not you. Disambiguation becomes the whole job.

Lean hard on the attributes that are unique: your exact address, your phone number, your founder's name, your founding year. Use sameAs aggressively, since it is the most direct tool available for saying "this one, not that one." And add Person schema for the owner — a named human is often the most distinguishing attribute a generically named business has.

The entity killers

Six problems account for most weak entities. They are listed roughly by how much damage they do, and none of them look like problems from the inside.

1. The ghost address

You moved. Forty directories did not. The machine now has two confidently stated locations for one business and no way to break the tie. This is the most destructive single item on the list because it is not merely incomplete — it is actively contradictory, and it persists for years unless someone deliberately hunts it down.

2. Call tracking sprawl

Different numbers deployed to different listings for attribution. Reasonable intent, damaging side effect: you have manufactured the exact inconsistency you would otherwise be trying to eliminate. If you use call tracking, use dynamic number insertion on your own site only and keep every external listing on the canonical number.

3. Name drift

The legal entity, the trading name, the name on the van and the name on the Google profile have all diverged slightly. Each variant looks harmless. Together they fragment one entity into four weak ones, none of which accumulates the corroboration it should.

4. Abandoned profiles

A Facebook page nobody has touched since 2019, listing hours you no longer keep. An old Yelp listing from a previous owner. Duplicate Google listings created accidentally and never merged. Each one is a source actively stating something false about you.

5. Blocked crawlers

All five layers can be perfect and still invisible if robots.txt blocks the systems that would read them. Many sites do this accidentally through a default template. It costs five minutes to check and occasionally explains everything.

6. Schema that contradicts the page

Markup claiming reviews, hours or content a visitor cannot actually see. Beyond being a guideline violation, it teaches the machine that your structured data is unreliable — which undermines the layer that was supposed to be your most trustworthy statement.

Run the audit yourself

Ninety minutes, a spreadsheet, and more patience than skill.

How long it takes

Realistic expectations, because this is slow work and knowing that up front prevents abandoning it at week three.

One closing thought. Everything in this article is available to any business of any size, costs almost nothing but time, and compounds rather than decaying. That combination is rare in marketing — and it is why we start client work here rather than with campaigns. Buying attention for a business a machine cannot verify is an expensive way to stay invisible.

Questions we get about this