Cosine similarity · Vector closeness measure
Cosine similarity is a measure of closeness between two vectors, based on the angle between them rather than their length. In GEO, it is used to compare a query’s embedding with that of a passage or page to estimate how close they are in meaning.
Standalone definition · 45 words
How is it calculated?
It equals the dot product of the two vectors divided by the product of their norms. The result lies between -1 and 1: the closer to 1, the more the vectors point in the same direction, so the closer in meaning the texts they represent.
How is it used in AI search and audits?
Vector search systems use it, or similar measures, to find the passages closest to a query. In an audit, you compute the similarity between the embedding of a page or chunk and that of a target query, as iPullRank does for its relevance scores.
What precautions apply?
A score is comparable only with the same embedding model, language and chunking. It measures semantic closeness, not quality or truthfulness: a page close to a query can still be weaker than another on authority or reliability.
Frequently asked questions
Is a similarity of 0.8 good?
There is no universal threshold. It is read by comparing pages on the same model and corpus.
Does Google use cosine similarity?
Google does not publish its internal measures. The term describes a general method, useful for modeling, not the formula of a specific engine.
Related terms
Further reading
- Profound vs Otterly vs Ahrefs Brand Radar: query fan-out comparison
- ← Back to the GEO glossary (114 terms)