Entities for GEO: Why Wikidata Is the Key to AI Visibility

"Apple" can mean a fruit at the grocery store or the technology company from Cupertino. For a human, context makes the difference obvious instantly. For a search engine or a language model, that requires something more concrete: an entity with a unique ID. A guiding principle up front: keywords describe words, entities describe things.

What entities actually are, technically

An entity has a unique identifier, a keyword is just a search string – that's the whole difference. Wikidata assigns every distinct concept a QID (Q plus number), for example Q180711 for search engine optimization, Q115564437 for ChatGPT, or Q807968 for Schema.org. These IDs form the backbone of the knowledge graphs that Google and large language models rely on to map terms precisely instead of just matching strings of letters.

When one word means several things: disambiguation in practice

• Apple: Q312 (Apple Inc.), Q89 (the fruit) or Q210593 (Apple Records) – a different entity depending on context
• Jaguar: Q26723 (the car brand), Q35694 (the animal) or Q221122 (Mac OS X 10.2) – without entity mapping the query stays ambiguous
• Golf: the VW Golf, the sport, or the Gulf of Mexico – the same string of letters, completely different concepts

Without an entity anchor, an AI can't cleanly cite a page because it has no proof of what's actually meant. With markup, it knows.

Keyword vs. entity: a direct comparison

Criterion	Keyword	Entity
What it is	A search string, a sequence of letters	A precisely identifiable concept with a unique ID
Ambiguity	High – the same string means many things	None – one ID means exactly one concept
Language dependence	One keyword per language	The same QID across all languages
Optimized for	Google SERP position	Citation in AI answers
Used by	Google, Bing, Ahrefs, Sistrix	ChatGPT, Perplexity, Gemini, Claude, Google AIO

Wikidata, Knowledge Graph and Entity Linking

Three concepts AI search builds on. Wikidata (Q2013) is the Wikimedia Foundation's free knowledge base and the de facto reference for public entities – Google, Bing, Meta and most AI systems sync against it. The Knowledge Graph (Q648625) is a network of entities and their relationships; the knowledge panel on the right of a Google SERP comes from exactly this graph. Entity Linking is the process of assigning a text mention to the correct entity: the sentence "Jaguar buys new factory" gets analyzed and the matching QID chosen. That's exactly what your sameAs tells the machine in advance, so it no longer has to guess.

What that looks like in JSON-LD

An article about search engine optimization with clean entity markup – copy the pattern, swap in your names and QIDs:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Entities for GEO",
  "author": { "@type": "Person", "name": "Knut Nickol" },
  "about": [{
    "@type": "Thing",
    "name": "Search Engine Optimization",
    "sameAs": "https://www.wikidata.org/wiki/Q180711"
  }],
  "mentions": [{
    "@type": "Thing",
    "name": "ChatGPT",
    "sameAs": "https://www.wikidata.org/wiki/Q115564437"
  }]
}

The load-bearing property is sameAs: it turns text mentions into machine-linkable entities. about marks the main topic, mentions secondary topics, author.sameAs and publisher.sameAs anchor the author and the brand.

Thesis vs. antithesis

The thesis: pack enough keywords into a text and you'll get found. The antithesis, backed by our own numbers: without clear entity mapping, a text stays ambiguous to AI systems – and ambiguous sources get cited less often than clearly linked ones. In our own tracker we currently count 617 Bing Copilot citations of rankmio.de over 60 days, a 5.4-fold increase versus the previous month. The most-cited page has a precise entity anchor in its JSON-LD, author markup with sameAs, and a body that answers the question in the first two sentences. Not luck, but a repeatable recipe.

What it comes down to

Keywords tell a search engine roughly what something is about. Entities tell it exactly what it's about. A principle worth keeping: entities without a citable body are invisible, a citable body without entities gets guessed at – both together get cited. For GEO, that precision is the difference between "might get found" and "gets cited as a source".

Sources

• Rankmio: Entities for GEO: Why ChatGPT Relies on Knowledge Graphs, Not Keywords (original article including disambiguation examples and tracker figures)
• Wikidata – the Wikimedia Foundation's free knowledge base, source of the QIDs
• Schema.org: sameAs – official documentation of the load-bearing property

FAQ

What is a Wikidata QID?

A unique identifier for a concept, person, company or term, in the form Q plus number – for example Q180711 for search engine optimization. Independent of language or spelling.

Is sameAs in Schema.org enough on its own?

It's a strong signal, but most effective combined with a citable body that answers the question directly, plus author and organization markup. Entities without a citable body stay invisible.

How do I find the right QID for a term?

Through the Wikidata search directly on wikidata.org. It's important to check the description so the entity truly matches the intended context, especially for ambiguous terms like Apple or Jaguar.

Do I need my own Wikidata entry for my brand?

Ideally yes. A minimal organization entry with sameAs to your website and social profiles is enough to anchor the brand as an entity. For authors, a LinkedIn URL is the minimum, your own QID the gold standard.

What do the 617 Bing Copilot citations actually mean?

That's the count measured in our own tracker of instances where Bing Copilot named rankmio.de as a source in a generated answer within 60 days.