THE SHORT ANSWER
An entity is a thing a search or AI system recognises as a distinct, identifiable object rather than a string of characters: a company, a person, a product, a place. Systems reconcile mentions across the web to decide which entity a page is about and what is true of it. When that reconciliation is weak, assistants blend your business with a similarly named one, or describe you with facts from a competitor's comparison page.
Search stopped being about strings and started being about things around 2012. Retrieval-based AI has pushed that much further, because a generated answer has to commit to a single referent: it cannot hedge across three companies with similar names the way a list of ten links quietly does.
That commitment is where most of the damage happens, and it is almost entirely fixable.
The numbers, at a glance
What an entity is: a distinct, identifiable thing, not a keyword string
How it is established: consistent naming, structured markup, and corroboration across independent sources
Most common failure: name variants across directories and social profiles, splitting one company into two entities
Second most common: no authoritative description on your own site, so the system borrows one
How a system decides which entity your page is about
Three inputs, roughly in order of weight. Explicit markup, where organisation schema names you and links your profiles. Consistency, where the same name, address, category and description appear across your site, your business profile, directories, registries and social accounts. And corroboration, where independent sources describe you the same way, which is what turns a claim into a fact from the system's point of view.
None of these is exotic. All three degrade quietly over years as an agency rebrands a listing, a new office address appears on one directory, or a social bio gets rewritten by someone in marketing who never saw the others.
What weak entity resolution looks like in practice
Blending. The assistant merges you with a similarly named company, attributing their services or their location to you.
Borrowed descriptions. Asked what you do, the assistant paraphrases a directory listing or a competitor's comparison page, because your own site never states it plainly in one place.
Stale facts. An old address, a discontinued service or a previous trading name persists, because the corrected version appears in only one place and the old one in six.
Absence. The assistant declines to say anything specific, which is what happens when signals conflict enough that nothing clears the confidence bar.
Fixing it: the consistency pass
Pick one canonical form of your company name and use it with no variation, including punctuation and legal suffix. Write one canonical two-sentence description of what you do, for whom, and where. Then propagate both, deliberately, to every place the web describes you: your own site footer and about page, organisation schema, your business profile, every directory listing, every social bio, every association or chamber membership page, and every partner site that lists you.
This is tedious and it is the single highest-return action available for entity strength. It also has a pleasant property: it is entirely under your control, unlike almost everything else in this discipline.
Building corroboration once consistency is fixed
Consistency stops the bleeding; corroboration builds the entity. What counts is independent sources describing you in the same terms: press mentions, industry association listings, review platforms with genuine volume, guest contributions with a proper author bio, and case studies published by clients rather than by you.
Two things that look like corroboration and are not. Syndicated press releases, which the systems recognise as one source duplicated. And directory listings you created yourself in bulk, which are consistency signals rather than independent testimony. The distinction matters: fifty self-created listings strengthen resolution and prove nothing, while three genuine third-party descriptions do both.
The entity audit
Choose one canonical company name and one canonical two-sentence description.
Search your own name and list every page in the first three results pages that describes you.
Correct each one to the canonical form, starting with the highest-authority sources.
Add or fix Organization schema linking your site to every official profile.
Ask three assistants who you are, and log which specific facts they get wrong.
Want leads like this in your pipeline?
Flock runs the campaigns, screens the enquiries and hands you only the ones that match your service area, job size and capacity. You pay per lead, not per month.
Book a 15-minute fit check | See lead package pricing
Related answers
Frequently asked questions
How long does entity correction take to show up?
Live-browsing assistants can reflect a corrected site within weeks. Anything baked into model training data waits for the next model release, which is months, so fix the source and be patient.
Do I need a Wikipedia page?
It helps considerably and is out of reach for most service businesses, since notability rules are strict. Consistent markup plus genuine third-party corroboration gets you most of the practical benefit.
Does my legal name have to match my trading name?
They can differ, but the relationship must be stated somewhere machine-readable, typically in organisation markup with an alternate name. Unexplained variants are what create two entities out of one company.
What if a competitor has a very similar name?
Lean harder on the distinguishing attributes: location, service area, founding year and category, stated consistently everywhere. Distinctiveness is what resolves a collision, not volume.
NEED A CLEARER PLAN?
Let’s turn your next move into momentum.
Talk to us →