A simple example
A keyword search for “cheap flight” may not match a page titled “low-cost airfare.” Semantic search can connect the two because the ideas are related. That is useful when users do not know the exact vocabulary used in your documents.
How it differs from keyword search
| Search type | Best at | Weakness |
|---|---|---|
| Keyword | Exact names, IDs, phrases, and filters | Misses related wording |
| Semantic | Intent and concept similarity | Can return something related but not actually correct |
| Hybrid | Combining exact constraints with meaning | Needs more design and evaluation |
The important boundary
Semantic similarity is not factual relevance. A document can be close in meaning and still be the wrong policy, version, customer, or date. Use metadata filters and human-readable source context when the answer has consequences.