Deutsch
Retrieval & grounding

Vector search

Vector search finds similar items by representing text, images or audio as vectors, called embeddings, and comparing the distance between them. It matches on meaning instead of on exact words.

How does vector search work?

An embedding model converts the content into vectors, which are stored in a vector database or a search index. The query is converted in the same way. The system returns the items whose vectors lie closest to the query vector.

IBM names Euclidean distance and cosine similarity as common measures of closeness, with cosine similarity ranging from -1 to 1. Exact comparison against millions of vectors is slow, so most systems use approximate nearest neighbor (ANN) algorithms. A widely used one is HNSW, introduced by Malkov and Yashunin in 2016, which builds a layered graph of vectors. Elastic notes that ANN gives up perfect accuracy for speed. Many systems combine vector scores with classic keyword scoring such as BM25, which is called hybrid search.

Embeddings: meaning as closeness in spaceSimplified two-dimensional sketch. Dog, puppy and cat sit close together, tax return and tax assessment far away in a group of their own. A query “young dog” sits between dog and puppy, and a circle marks its nearest neighbors. Real embeddings often have 256, 512 or 1024 dimensions.dimension 1dimension 2large distancedogpuppycattax returntax assessment“young dog”embedding of a termquery as an embeddingnearest neighborsClose together meanssimilar meaning.Far apart meanslittle in common.Real embeddings oftenhave 256, 512 or1024 dimensions.Made-up example in 2 dimensions, positions not calculated.
Made-up example. According to Google, points that sit close together in embedding space are more closely related than distant ones. Real embeddings often have 256, 512 or 1024 dimensions. Sources: Google Machine Learning Crash Course: Embeddings; Google Machine Learning Crash Course: Embedding space and static embeddings.

Tiger Data describes vector search as a building block of semantic search. Vector search is the technique that compares embeddings. Semantic search is the broader goal of returning results that fit the intent of a query, which can use further methods such as knowledge graphs (assessment based on that source).

What does it mean for SEO?

Retrieval-augmented generation (RAG) uses vector search to find passages that fit a question. Clearly structured pages with self-contained sections are easier to retrieve this way (assessment). No documentation confirms vector search as a ranking method in Google Search.

Retrieval-augmented generation: retrieve first, then answerA question triggers a search in a search index or vector database. The best passages are added to the prompt together with the question. The language model writes the answer from them and points to the sources of the passages. The answer can only cite what was retrieved first.1 QUESTION2 RETRIEVAL3 CONTEXT4 ANSWERQuestionfrom a useror an appRetrievalsearch orvector search findsmatching passagesKnowledge sourcesearch index, web,vector databasePrompt with contextquestion + passages[1] passage · URL A[2] passage · URL B[3] passage · URL CLLMAnswertext withsource links[1][3]Citations point to retrieved passagesSchematic, based on Lewis et al. (2020) and Google’s grounding docs.
In retrieval-augmented generation, the system first retrieves matching passages and passes them to the language model as context. Google calls the link between model output and verifiable sources grounding. The answer can only cite what was retrieved first (an inference from the method, not a Google statement). Sources: Lewis et al. (2020): Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (arXiv 2005.11401); Google Cloud: Grounding overview (Vertex AI); Google Cloud: Grounding with Google Search (Vertex AI).

Example

A search index holds a page called “Washing a wool sweater by hand”. The query “stop a cardigan from shrinking” contains none of its words but sits close to it in vector space. Vector search returns the page, a pure keyword search would not.

FAQ

What is the difference between vector search and keyword search? Keyword search compares terms. Vector search compares meaning through embeddings.

Do you need a vector database for it? Not always. Several search engines offer vector features inside their own index.

Sources

  1. What is vector search? - Elasticelastic.co
  2. What is vector search? - IBM Thinkibm.com
  3. Efficient and robust approximate nearest neighbor search using HNSW graphs - arXiv:1603.09320 (Malkov, Yashunin)arxiv.org
  4. Vector search vs semantic search - Tiger Data (secondary source)tigerdata.com

Last updated: