The Entity Contradiction Audit

When an AI assistant declines to say anything about a business, the usual assumption is that it has never heard of it. Often the opposite is true. It has heard several things, they disagree, and it has no basis for choosing between them.

That is an entity contradiction, and it is the most common fixable cause of invisibility I find.

why does AI give vague answers about my business?Usually because your facts disagree with each other across the web. A model that finds two different phone numbers for you reports neither — hedging is what uncertainty looks like from the outside.

What an entity is, in one paragraph

To a machine your business is not a website. It is an entity: a thing with properties — a name, an address, a phone number, a category, a set of people, a set of profiles elsewhere. Those properties get assembled from everywhere the business appears. When the sources agree, the entity resolves cleanly and a model will state things about it. When they disagree, confidence drops, and the practical result is hedging or omission.

Nobody tells you this is happening. You simply do not come up.

Running the audit

This takes about ninety minutes and needs a spreadsheet, not a tool.

Column one: every place your business appears. Your website, your Google Business Profile, LinkedIn, Yelp, Bing Places, Apple Business Connect, your chamber of commerce, every industry directory, every aggregator you never signed up for, your social profiles, your email signature, your invoices. Search your business name and your phone number in quotes and work through the results. Most businesses find between fifteen and forty.

Columns two onward: the facts. Exact business name. Street address including suite. Phone. Website URL. Primary category. Hours. Owner or principal name.

Then read down each column. Every place a column is not identical is a contradiction.

Record what is there, not what should be there. The instinct is to write the correct value in every row. The audit only works if it captures what each source actually says, including the wrong ones.

The contradictions that cost you most

Ranked by how much damage they do, which is not the order people fix them in.

1. Two phone numbers. The single most damaging. Usually an old number on a directory nobody remembers claiming, or a tracking number from a former agency. A model finding two numbers for one business has no way to choose and will often give neither.

2. Two business names. “Acme Dental” and “Acme Dental Care LLC” and “Dr. Smith — Acme Dental” may be the same to you. To a machine they are three candidate entities. Pick one legal-but-usable form and make everything match it exactly, down to punctuation.

3. Two addresses. Usually a suite number present in some places and missing in others, or a former location still listed somewhere. For a service-area business, a public address in some places and a service area in others is the same problem.

4. Two websites. An old domain still live, or a www and non-www version both resolving without a canonical. A business pointing at two domains is a business a machine cannot confidently attach to either.

5. Conflicting hours. Lower stakes but still a confidence signal, and the easiest to let rot.

6. Category drift. “Marketing consultant” in one place, “advertising agency” in another, “business consultant” in a third. This does not block an answer so much as blur which question you are the answer to.

Fixing, in the right order

First, decide the canonical version. One name, one address format, one phone, one URL, one primary category. Write it down. This is the only creative decision in the process; everything after is clerical.

Then fix your own site. It is the source others copy from and the one a machine treats as most authoritative about you. Make the visible text and the structured data agree with your canonical version, and make them agree with each other.

Then your Google Business Profile, which feeds more downstream systems than anything else you control.

Then the big aggregators, because many small directories pull from them. Fixing an aggregator can correct a dozen sites you have never visited.

Then the long tail, worst first. You will not clear all of it. You do not need to — you need the weight of agreement to be overwhelming.

Then declare the relationships. Use sameAs in your structured data to list every profile you control. This is the step almost everyone skips, and it is the one that tells a machine these scattered listings are the same entity rather than similar ones.

What success looks like

Not a score. Ask an assistant who runs the business, where it is, what it does, and how to reach it. Correct and specific answers mean the entity resolved. Hedging, confusion with a similarly named business, or a flat refusal means a contradiction survives.

Re-run it monthly for a while. Directories regenerate bad data, and a business that was clean in March can be contradictory by June because an aggregator refreshed from a stale source.

This work is tedious and it is not clever. It is also the highest-yield thing most businesses can do for AI visibility, because it removes the specific obstacle that stops a model talking about you at all.

Find out what is actually being said about you. The free audit asks a live assistant real questions about your business and reports which sources shaped the answer.

Run the free audit

Common questions

What is an entity contradiction?
It is when the facts about your business disagree across the places it appears — two phone numbers, two versions of the name, an address with a suite number in some listings and without it in others. AI systems assemble an understanding of a business from many sources. When those sources conflict, confidence falls and the practical result is that the model hedges or omits the business rather than stating something it cannot verify.

How do I find contradictions about my business?
Build a spreadsheet listing every place the business appears — your site, Google Business Profile, LinkedIn, Yelp, Bing Places, Apple Business Connect, chambers, industry directories, aggregators and social profiles. Search your business name and your phone number in quotes to find listings you never created. Record the exact name, address, phone, website, category and hours each source actually states, then read down each column. Any column that is not identical everywhere is a contradiction.

Which contradictions matter most?
Conflicting phone numbers do the most damage, because a model finding two has no basis for choosing and often gives neither. Then conflicting business names, since variants read as separate candidate entities. Then addresses, particularly a missing or inconsistent suite number. Then multiple live websites. Conflicting hours and drifting categories matter less but still reduce confidence.

In what order should I fix them?
Decide your canonical version first — one name, address, phone, URL and primary category. Fix your own website next, including making visible text and structured data agree. Then your Google Business Profile, then the major aggregators since smaller directories copy from them, then the long tail worst first. Finally declare the relationships with sameAs in your structured data so a machine knows those profiles are the same entity.

How often should this be re-checked?
Monthly at first. Directories regenerate data from stale sources, so a business that was consistent in one quarter can become contradictory in the next without anyone touching it. The practical test is to ask an assistant who runs the business, where it is and how to reach it: correct, specific answers indicate the entity is resolving, while hedging or confusion with a similar business indicates a contradiction remains.

Your facts disagree with each other. That is why you are invisible.