Skip to main content
AI & Agents

Semantic Search

Semantic search finds results by meaning rather than by matching words, so a query about "reducing customer churn" can surface a document about "improving retention" even though they share no vocabulary. It works by comparing embeddings — numerical representations of meaning — instead of comparing text.

The gain over keyword search is recall on paraphrased queries, which is most of how people actually type. The corresponding weakness is precision on exact strings: part numbers, error codes and clause references are exactly where embedding similarity performs worst, and exactly what enterprise users search for most.

That is why production systems rarely use semantic search alone. Hybrid retrieval runs keyword and vector search together and reranks the combined results, which captures conceptual matches without losing exact ones. A retrieval system that underperforms is very often a pure-vector system that should have been hybrid.

Codazz builds this in production — RAG Development.

FAQ

Semantic Search
FAQ.

Common questions about semantic search.

Ask Us Anything

Better at paraphrase, worse at exact strings. Semantic search finds conceptually related content that shares no words; keyword search reliably finds a specific part number or error code. Production systems combine both and rerank the merged results, because each covers the other's failure mode.

Related terms

Embedding

An embedding is a list of numbers representing the meaning of a piece of content, produced by a model so that semantically similar items sit close together in the resulting space.

Vector Database

A vector database stores text, images or other content as numerical embeddings and retrieves items by semantic similarity rather than exact keyword match.

RAG (Retrieval-Augmented Generation)

RAG is a technique where an AI system retrieves relevant passages from your own documents at query time and generates an answer grounded in them, with citations.

Browse the full glossary →