← All terms

Semantic Search

Searching by the meaning of text rather than by exact keyword match. The foundation of modern RAG systems.

What is semantic search?

Semantic search finds text by meaning, not by exact word match. For the query "How do I cancel my subscription?" it can also return a document that describes "terminating a payment plan." Classic full-text search would miss this document because it contains neither "cancel" nor "subscription."

How semantic search works

1. Documents are converted into vectors by an embedding model and stored in a vector database.
2. The user's query is turned into a vector by the same model.
3. The database finds vectors that sit closest to the query vector (e.g. using cosine similarity) - i.e. documents with similar meaning.

Why semantics wins

• Handles synonyms, translations, and varied phrasing
• Understands context - "apple" the fruit vs. Apple the brand
• Works even for natural-language questions
• It is the backbone of RAG systems and modern chatbots

Limits of semantic search

For short, exact queries (catalog numbers, SKUs, names), pure semantic search may fail. The fix is hybrid search - combining semantics with keyword matching, often paired with re-ranking.