Semantic Search
Search that matches on meaning rather than exact words, so “car” can find “automobile.”
Think of it like
A helpful librarian who understands what you’re after, not a card catalog that only matches the exact title.
Example
You search “how do I cancel my plan” and it finds the doc titled “Ending your subscription,” despite zero shared keywords.
How it actually works
Semantic search turns both the query and the documents into embeddings and finds the ones closest in meaning-space, so it catches paraphrases and synonyms that keyword search misses. The flip side is that it can miss exact matches — a specific error code or product name — because those are about the literal string, not the vibe. That weakness is exactly why it’s so often paired with keyword search.
For product teams
Semantic search makes finding things feel intuitive, but it can fumble exact terms — rarely the whole answer alone.
For engineers
Embed query and documents, then rank by vector similarity in the embedding space.
Related
- Embeddings — What powers meaning-based matching.
- BM25 — The word-matching approach it complements.
- Hybrid Search — Combining both approaches.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome