Entity Linking, and Why AI Needs to Know Who You Are
Entity resolution decides that several mentions are the same thing. Entity linking goes one step further: it connects a mention in text to a known record of that thing in a knowledge base. When an article says "the Azure" and a system attaches that phrase to a specific entry with a stable identity, an address, a set of facts and a history, that is entity linking. It is how a word on a page becomes a thing a machine knows about.
The difference in one line
Resolution asks "are these the same?" Linking asks "which known thing is this?" The first clusters mentions together. The second anchors them to an identity that exists independently of any single page.
How it works
Linking runs in three steps.
First, the system spots the mentions. In "we stayed at the Azure near Walking Street and ate at Salt", it notices that "the Azure", "Walking Street" and "Salt" are probably names of things rather than ordinary words.
Second, for each mention it generates candidates from its knowledge base. "Azure" could be a hotel, a cloud platform, a colour, a yacht. "Salt" could be a restaurant, a mineral, a band.
Third, it uses the surrounding context to pick one. "Stayed at", "near Walking Street" and "ate at" make the hotel and the restaurant the obvious choices. The mention is then tied to the matching entry, usually by an identifier rather than a name, so that any future mention resolving to the same entry inherits everything already known about it.
Where you have seen it
Wikipedia's blue links are entity linking done by hand. Google's knowledge panel is entity linking done at scale: when you search for a business and a card of facts appears, the engine has linked your query to a specific entity in its graph. AI assistants now do the same thing invisibly while answering, pulling the facts attached to the entity rather than re reading every page that mentions it.
Why it matters for AI answers
An assistant that can link "the Azure" to your entity gets to use everything attached to that entity: your verified address, your amenities, your hours, your reviews, your relationship to your restaurant and spa. An assistant that cannot link it is stuck with whatever is on the single page it happens to be reading, which is usually incomplete and sometimes wrong.
Linking also protects you from confusion. If "Azure" links cleanly to your entity, facts about the cloud platform or the hotel in another country stay with their own entities. Without a clean link, the model is left to guess from context, and context is often thin.
What makes you linkable
A machine can only link to an entity that exists in a knowledge base and is distinguishable from its neighbours. Three things help.
Have a canonical home. One page, usually on your own site, that is unambiguously about your business and nothing else, with your full name, address, contact details and a plain description. This is what other systems point to.
Be present in the public graphs. A Google Business Profile, a Wikidata entry where appropriate, and consistent listings on major directories give linkers a record to land on. Each should carry the same name and the same details.
Declare the link yourself. Structured data on your site can say "this page describes this entity" and include a sameAs pointer to your profiles elsewhere. You are telling the machine which record you are, instead of hoping it works it out.
Disambiguate on the page. If your name is shared with other things, make the context obvious in the first lines: what you are, where you are, who you belong to.
The takeaway
Entity linking is how a machine turns your name on a page into you. The stronger and more consistent your identity across the places machines read, the more reliably every mention of you connects back to the right record, with all your facts attached.