AI & Infrastructure
Entity Linking

Entity Linking: Connecting Names to Knowledge Base Entries

Entity linking is the NLP task that connects a name in text, such as “Apple”, to the one knowledge base entry it refers to. It decides whether the mention means the company, the fruit or something else.
July 3, 2026
Cihan Geyik
Table of Content

A prompt asks: “Is Monday good for agency project management?” A person reads “Monday” as monday.com, the work management software. A machine has to work that out. Entity linking is the step that ties a name in text to one entry in a knowledge base, so the system knows which thing it’s dealing with. Get it wrong and every fact that follows belongs to someone else.

Entity linking turns a string into an ID. Once a recognizer flags “Monday” as a name, the linker matches it to one knowledge base entry, or decides that no entry fits.

How does entity linking work?

Detect the mention → Generate candidates from the knowledge base → Score each candidate in context → Pick one ID or return no match
  1. Mention detection. Entity recognition finds the span “Monday” and tags it as a likely organization.
  2. Candidate generation. The linker looks up every entry whose name or alias matches, such as the weekday and the software company.
  3. Disambiguation. Each candidate is scored against the context words, here “agency” and “project management.” Prior popularity counts too.
  4. Decision. The top candidate wins if it clears a confidence threshold. Otherwise the mention gets a NIL label, meaning no match.
Mention: “Monday” (context: agency project management)
Candidate: Monday, day of the week → weak fit with the context
Candidate: monday.com, work management software → strong fit
Fallback: NIL → used when no candidate clears the threshold

Modern linkers do this with neural models. The BLINK paper by Wu and colleagues (2019) uses two stages. A bi-encoder embeds the mention and each entity description, then retrieves candidates by vector search. A cross-encoder reads mention and candidate together and re-ranks them. Each entity needs only a short text description, so the model can link to entities it never saw in training.

Entity linking vs entity recognition vs entity resolution

These terms get used loosely, and some glossaries treat them as synonyms. They answer different questions.

TaskQuestion it answersOutput
Entity recognitionIs this span a name, and of what type?“Monday” = organization
Entity linkingWhich knowledge base entry is it?The ID of monday.com
Entity resolutionAre these two records the same thing?Merged duplicates in a CRM or catalog
Coreference resolutionWhich later words refer back to it?“the company” and “it” tied to monday.com

The types a recognizer assigns, such as person, organization or product, are covered on our named entities page.

Where does entity linking show up in search and AI?

Google’s Cloud Natural Language API exposes it directly. For many entities it returns a Wikipedia URL and a Knowledge Graph ID, called a MID, next to the name and type. That’s entity linking offered as a service.

spaCy, a popular open-source NLP library, ships an EntityLinker component. Its default candidate lookup matches exact, case-dependent aliases. If no candidate scores above the threshold, the mention is labeled NIL. Small details like capitalization decide whether a brand gets linked at all.

AI answer engines keep their linkers private. Yet every answer about a named company depends on the step, because facts come from the entry the name was linked to in a knowledge graph or index. When an answer blends two companies that share a name, a linking error is the likely cause.

Why entity linking matters for brands

  • Namesakes. When context is thin, linkers favor the entity a name most often means. A smaller company with a famous namesake loses that tiebreak.
  • New brands. A company with no knowledge base entry gets NIL. The system then has only the page text to go on.
  • Rebrands. Old aliases keep pointing to the old record until sources update.
  • Common-word names. Products named after ordinary words need stronger context than coined names.
  • Markets. A name that is unique in one country can be a common word or another company elsewhere.
Context is the part you control. Linkers score the words around a mention, so the category, product and integration names near your brand name do real work.

How to make your brand easier to link

  1. Pair the name with its category on first mention, as in “Example, the invoicing tool for freelancers.”
  2. Use one spelling and capitalization everywhere, since simple alias lookups are exact.
  3. Give knowledge bases an entry to link to where you qualify. Our page on brand entities covers that setup.
  4. Mention the entities you relate to, such as your category, integrations and rivals. They are the context a linker scores.
  5. Keep former names on your site after a rebrand, so old aliases still lead to you.

For the wider program around these fixes, see our guide on how to improve brand AI visibility.

How Ansvisor helps

Linking errors leave traces in the sources an answer cites. Ansvisor’s Citations Monitoring lists the URLs each AI platform cites for every tracked prompt. If answers about your brand cite a namesake’s website or its Wikipedia page, the engine has likely linked your name to the wrong entity. Track branded prompts in every market you serve to catch the countries where that happens.

Named Entity Linking, Entity Disambiguation, Named Entity Disambiguation, Wikification, NEL

Let AI Search Work for Your Growth

Each prompt you win brings more mentions, more citations and more buyers. Ansvisor keeps finding the next one.

AI Answers Mentions Citations Visibility Score AI Share of Voice AI Traffic Revenue
Win Back Search VisibilityTry Ansvisor 14-day Free →✓ Zero risk. Cancel anytime.

FAQ

Frequently asked questions.

What is Entity Linking?

Entity linking is the step that connects a name in text to one entry in a knowledge base, such as Wikidata or Google’s Knowledge Graph. It decides which real thing a mention refers to. If no entry fits, a good system returns no match instead of forcing a wrong one.

Why is Entity Linking important?

Every fact a system pulls about a name depends on picking the right entity first. A wrong link attaches another company’s founders, prices or reviews to yours. For brands that share a name with something better known, linking errors are a common cause of confused AI answers.

How does Entity Linking work?

A recognizer first finds the name in the text. The linker then pulls candidate entries whose names or aliases match. It scores each candidate against the surrounding words and how often the name usually means that entity. It picks the best match, or returns no match when nothing scores high enough.

What is the difference between entity linking and entity recognition?

Entity recognition finds the span of text that is a name and gives it a type, such as organization or person. Entity linking takes that span and decides which specific entry it means in a knowledge base. Recognition comes first, and linking depends on it.

Which tools help analyze Entity Linking?

spaCy ships an EntityLinker component that matches mentions against a knowledge base you supply. Google’s Cloud Natural Language API returns a Wikipedia URL and Knowledge Graph ID for many entities. To spot linking errors in AI answers, Ansvisor shows which sources each platform cites for your brand prompts.

Help us grow the AI Visibility Glossary

New terms are added regularly.

Help us improve the page or suggest a new term →
About the Author
Cihan Geyik

Cihan Geyik

Co-founder at Ansvisor

Cihan Geyik is the co-founder of Ansvisor, an open-source AI Visibility platform for AI Search. With more than 15 years of experience in digital marketing and growth, he writes about AI visibility, AI search, AEO, GEO, citations, and answer engines. He focuses on helping brands understand and improve their presence across ChatGPT, Gemini, Google AI Overviews, Bing AI Mode and other AI answer engines.

Google Preferred Sources
Get more Ansvisor insights in Google

Add Ansvisor as a preferred source and Google will show you more of our AI search, AEO and GEO guides.