Research
entities, not keywords: how machines decide who you are
Ask an AI about a company that shares its name with three others and watch it hedge, blend or pick the wrong one. That failure is called entity resolution, and it decides something prior to ranking: whether the machine knows which thing you are at all. Modern engines do not index strings of keywords, they map entities, distinct things with names, properties and relationships, and your business is one of them, well-defined or not.
what an entity is to a machine
An entity is a node in a knowledge graph: this company, founded by these people, offering these services, located here, same as this LinkedIn page and this directory listing. Engines assemble that node from every mention of you they can find, and the assembly is only as good as the agreement between sources. Where descriptions align, confidence grows. Where they conflict, the machine hedges, and hedged entities get recommended less.
consistency is the mechanism, not a nicety
The practical work is unglamorous: the same business name, description, address and offering everywhere you appear, structured data on your own site declaring who you are, explicit links between your profiles, and a description phrased the way you want machines to repeat it. Every consistent mention is a vote for the same node. Every variant spelling, stale listing and contradictory bio splits the vote.
the shared-name problem
If your brand name is a common word or is shared across industries, disambiguation is your first battle, and context is the weapon. Machines separate same-named entities by the company they keep: the services, founders, locations and domains that co-occur with the name. Anchor every mention with that context, consistently, and the graph learns which node the words belong to. This is also the honest argument for distinctive positioning language: a phrase only you use resolves instantly.
the honest catch
Entity work is slow and mostly invisible, with no dashboard cheering you on, and it cannot be finished, only maintained. It is also unusually durable: keyword tactics age with algorithms, while a well-defined entity survives platform shifts because every engine needs to know what things are. Quiet work, compound interest.
Common questions
Where does structured data fit in?
- Schema markup on your own site is you declaring your entity's properties in machine-readable form: organization, people, services, and links to your other profiles. It is the cheapest strong signal available.
We share a name with other companies. Priorities?
- Context everywhere: name plus what you do plus where, in every bio, directory and byline. Never let the bare name travel alone.
How do we check how machines see us?
- Ask the assistants who you are and watch for blending with the other name-holders. Confusion in the answer is a to-do list for your consistency work.

Tom Claydon
Co-founder of Nudge
Ask Tom anything about search, he answers on WhatsApp.