# Semantic search

> Semantic search interprets the meaning and intent behind a query rather than matching exact keywords, so a search for "warm jacket for hiking" can return the right product even when those words aren't in its title.

_Source: https://nobi.ai/glossary/semantic-search_

## What does "semantic search" mean?

Semantic search is a search approach that interprets the **meaning and intent**
behind a query rather than scanning for exact word matches. A shopper who types
"warm jacket for hiking" can reach a product titled "thermal trail shell"
because semantic search reads what the person means, not just the words they
typed.

It is the opposite of traditional keyword search, which only returns results
when the shopper's words appear in your product titles, descriptions, or tags.

## How does semantic search work?

Semantic search converts text into **numerical representations called
embeddings** that capture meaning, then finds results whose embeddings are
closest to the query's. Because similarity is measured by meaning, synonyms,
paraphrases, and descriptive phrases all map to the right products without
anyone hand-writing synonym rules.

In practice that means:

- "running shoes for flat feet" can surface stability shoes even if "flat feet"
  never appears in a title.
- Misspellings and alternate phrasings still resolve to the right items.
- New phrasings work without a merchandiser adding a rule for each one.

## Why does semantic search matter?

When search only matches keywords, every query a shopper phrases in their own
words risks a dead-end "no results" page - and most shoppers who hit one simply
leave. Semantic search closes that gap, which is why it tends to move the
metrics that matter: a lower zero-result rate and a higher conversion rate on
search sessions.

It also reduces ongoing work. Teams running keyword search spend time writing
synonym lists and pinning fixes for queries the engine gets wrong; meaning-based
matching removes much of that maintenance.

For stores using [Nobi](https://dashboard.nobi.ai), semantic matching is the
default: a shopper's natural-language query is matched against the live catalog,
so descriptive searches resolve without manual synonym upkeep.

## Frequently asked questions

**What is the difference between semantic search and keyword search?**
Keyword search matches literal words against your catalog, so it fails when
shoppers describe a product differently than you labeled it. Semantic search
interprets meaning and intent, so descriptive and long-tail queries still
return relevant results.

**Does semantic search replace filters and facets?**
No. Semantic search handles how shoppers describe what they want; filters let
them narrow by attributes like size, color, and price. The two work together.

**How do you measure whether semantic search is working?**
Watch the zero-result rate (it should fall) and the search-to-cart or
conversion rate on search sessions (it should rise).