Definition
Vector search finds information by meaning rather than exact words, by comparing numerical representations (embeddings) of text.
It powers the retrieval step in most RAG systems. In practice it works best combined with metadata filters and keyword signals, so results are both relevant and permitted.
Why it matters
Keyword search fails when people use different words for the same thing: a customer asks about a refund and the policy says reimbursement. Vector search matches by meaning, so it finds the relevant passage even when the wording differs.
It is not a complete answer on its own. Exact identifiers such as invoice numbers still need keyword search, and permissions must be applied as filters so people only retrieve what they are allowed to see.
An example
A support assistant receives the question: my parcel never showed up. Vector search retrieves the lost-shipment policy, even though it never uses the words parcel or showed up. A metadata filter limits results to the customer's country, so the assistant quotes the right policy.
Illustrative example, not a client project.
Common questions about Vector search
What is an embedding?
A list of numbers that represents the meaning of a piece of text. Texts with similar meanings have embeddings that are close together, which is what vector search compares.
Do I need a special database for vector search?
Not always. Dedicated vector databases exist, but many relational and search databases now support vector indexes, which can be simpler to run.
Read more
Related terms