Tools & retrieval
What is Semantic Search?
Semantic search ranks items by embedding similarity instead of exact keywords.
Similarity is useful for discovery, not proof. Retrieved neighbors still need verification.
Why it matters
Semantic search finds documents by meaning, not just keywords. When a user asks "how do agents handle errors" and the document says "failure recovery in agentic systems," keyword search misses the match but semantic search finds it.
Semantic search uses embeddings — vector representations of text — to compute similarity. It powers the retrieval stage of RAG systems.
Key takeaways
- 1Semantic search matches by meaning, not keywords.
- 2It uses embeddings to compute similarity between queries and documents.
- 3Similarity is useful for discovery, not proof — retrieved results still need verification.
Common mistakes
- ✕Trusting semantic similarity as truth — similar vectors can be semantically unrelated.
- ✕Using semantic search without a re-ranking step.
Related terms
Concept neighborhood
Terms linked from Semantic Search in the glossary graph.
- Semantic Search
- Embeddings
- Vector Database
- Retrieval