Semantic search
Definition
Semantic search is a way of searching by meaning instead of exact words, so it finds passages that say what you asked even when they use different wording. It usually works by comparing embeddings, which are lists of numbers that capture what a piece of text means.
By Ronan Liedmeier, fractional head of growth and course author · Published 17 October 2026 · Last updated 17 October 2026
Also called: vector search, meaning-based search
People rarely phrase a question the way the answering document is written. A sales rep types "can we bill quarterly" while the contract says "payment every three months", and a keyword search comes back empty.
How it works
Documents are split into passages, and an embedding model turns each passage into an embedding. Your question gets the same treatment. The search, often run on a vector database, returns the passages whose numbers sit closest to the question's. Closeness in that number space follows closeness in meaning, so paraphrases and synonyms land near each other.
Many retrieval systems pair it with plain keyword search, called hybrid search, and then rerank the top results with a second model. Each method catches matches the other one misses.
Example
A marketing analyst at a software company keeps three years of win/loss call notes in a shared drive. She searches "prospects who left because setup took too long". Semantic search returns notes that say "onboarding dragged on for a quarter" and "IT never finished the integration", and neither contains the word setup.
Common mistake
Using semantic search alone for lookups built around exact identifiers. Matching by meaning is weak on SKUs and error codes, and a search for "E-4021" can return a passage about a similar code. Pair it with keyword search for those. Questions about numbers in a table, like "spend by campaign last week", are a database query and belong in SQL.
Learn it in AI Academy
- The Research SystemExpert · 8 min
- RAG: When It's the Right AnswerExpert · 9 min
Related terms
- Embedding
- An embedding is a list of numbers that represents the meaning of a piece of text, so a computer can measure how similar two texts are. Texts about the same idea get similar numbers even when they use different words.
- Vector database
- A vector database is a database built to store embeddings, the number lists that represent the meaning of text, and to quickly find the ones closest to a new question. In many RAG systems it is the place where the search step happens.
- Retrieval-augmented generation (RAG)
- Retrieval-augmented generation (RAG) is a way of getting an AI model to answer from your own documents. A search step first finds the passages that match the question, then the model writes its answer using those passages.
- Chunking
- Chunking is splitting long documents into smaller pieces before they go into a RAG system, so a search can find and return just the relevant part. Where you cut the text decides whether each piece still makes sense on its own.