Semantic search
Semantic search finds documents by meaning rather than keyword match, using embedding similarity in a vector space.
Traditional keyword search (BM25, TF-IDF) matches strings. Semantic search embeds the query and corpus into vectors and ranks documents by cosine similarity. The result: queries like 'how do I cancel my plan' match documents about 'subscription termination' even with no shared words. Production stacks in 2026 almost always use hybrid search — semantic plus keyword — and re-rank the top-K with a separate model, because pure semantic search misses exact-match terms (product SKUs, error codes).
When to use semantic search
- Search over long-form content.
- RAG retrieval.
- Recommendation by similarity.
Common mistakes
- Skipping hybrid + re-rank — pure semantic search misses exact terms.
- Re-embedding the whole corpus on every prompt change — only the query side changes.
FAQ
What is semantic search?
Semantic search finds documents by meaning rather than keyword match, using embedding similarity in a vector space.
When should I use semantic search?
Search over long-form content. RAG retrieval. Recommendation by similarity.
What are the most common mistakes with semantic search?
Skipping hybrid + re-rank — pure semantic search misses exact terms. Re-embedding the whole corpus on every prompt change — only the query side changes.
Related terms
- Embeddings — Embeddings are dense numeric vectors that represent the meaning of text, images, or other data, allowing similarity to be measured as vector distance.
- Vector database — A vector database stores embeddings and performs approximate nearest-neighbor search at scale, the persistence layer behind RAG and semantic search.
- Retrieval-augmented generation (RAG) — Retrieval-augmented generation (RAG) injects relevant documents into the prompt at query time so the model answers from your data instead of its training memory.
Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/semantic-search.md.