Nobody knows who you are: the entity problem behind most AI invisibility

A client asked us why ChatGPT kept recommending a competitor with a worse website. We asked ChatGPT what the client's company did. It described a different business, in a different county, with a similar name. Then we asked about the competitor and got an accurate paragraph: services, location, founder, founding year, a couple of review scores. The competitor's website was worse. Its identity was not.

This is the entity problem, and once you see it you find it everywhere.

Engines recommend entities, not pages

Classic search matched a query to a page. AI engines do something different. When a user asks for a plumber in Bexhill, the engine assembles candidate businesses from everything it knows: websites, Google Business Profile data, Bing, directories, review platforms, LinkedIn, press, Companies House, Reddit. For each candidate it is effectively asking: am I sure who this is, what they do, where they are, and that these twelve mentions all refer to the same firm?

When the answer is yes, the business becomes a thing the engine can hold facts about and recommend. When the mentions disagree, it either fails to connect them (so your reviews on one platform do not count towards the company described on another) or it hedges and picks a candidate it is sure about. Confidence, not quality, decides the shortlist.

How businesses fragment themselves

None of this is deliberate. It accumulates:

• The website says "Smith & Co Plumbing". Google Business Profile says "Smith and Co Plumbing & Heating Ltd". LinkedIn says "Smith Plumbing". Companies House says "SMITH PLUMBING SERVICES LIMITED".
• Three different phone numbers across old directory listings, two of them dead.
• A description rewritten for each platform because "each audience is different".
• An About page with no names, no founding year and no address because the founder works from home.
• A founder whose LinkedIn headline mentions a previous employer.
• Organization schema generated by a plugin with a placeholder name, or none at all.
• A trading name that is also the name of a shop in Bexhill, Massachusetts.

Each is trivial. Together they mean the engine is looking at a cloud of partial matches rather than one business, and a cloud does not get recommended.

The symptoms

Ask the four main engines "What is [company] and what does it do?", "Who founded it and when?" and "Where is it based?" If the answers are inconsistent between engines, or vague ("appears to be a marketing company"), or wrong, or borrowed from a namesake, that is the entity problem. A second symptom is subtler: your third-party mentions never show up in answers about you. The engine has not linked them to you, so they do nothing.

What resolving an entity actually involves

It is consistency work, and it is dull, which is why it does not get done.

One canonical description, about 25 words, stating name, what, for whom, where. Used verbatim everywhere: About page, LinkedIn, Google Business Profile, Bing Places, Apple Business Connect, directories, guest bios, press boilerplate.

An About page that reads like a fact sheet at the top: legal name and trading name, founders, founding year, who runs it now with photos and roles, full address, areas served, registration number, accreditations. Then the story.

Organization or LocalBusiness schema on the homepage with a stable @id, the canonical description, address, phone, founders, and sameAs links to every official profile. sameAs is the explicit instruction "these are all me". Person schema for the founders, linked back with worksFor.

Every profile you control edited to match: same name format, same address format, same number, same description, same category. A spreadsheet with a date-checked column. Dead listings claimed and corrected or removed.

Third parties asked to use the canonical description when they mention you.

Then the test again, quarterly. The answers get more accurate as the sources align, and the third-party mentions start appearing in them.

Why this comes before content

We get asked for content plans by businesses whose entity is fragmented, and we push back. Content is fuel; the entity is the vehicle. A brilliant guide published by a business the engine cannot confidently identify earns a citation for the page and nothing for the company. The same guide published by a resolved entity lifts every future answer about that company. Two weeks of consistency work, before a single new page, is the highest-return task in AI visibility for most small and mid-sized businesses.

The competitor with the worse website, by the way, had done exactly that work. Their agency had a checklist. Now so does our client.