← Back to the A-Z glossary
AI glossary · P1

Semantic search

Semantic search finds the meaning of a question, even when the exact words change.

RAG 4 min read Updated 2026-05-22
— Definition

Semantic search, in plain words

Semantic search finds the meaning of a question, even when the exact words change.

Present semantic search as search by meaning, not by exact words.

A concrete example

I search 'how to reduce customer attrition.' Semantic search finds a doc that talks about 'fighting churn.' A keyword search misses it.

Why it matters

It's the foundation of every modern RAG. Without semantic search, your assistant misses half the answers.

You run into it in RAG, embeddings, vector databases, and internal search engines.

Don't mix it up with

Embedding: An embedding turns a piece of text into a list of numbers so you can compare its meaning with other texts.

Vector database: A vector database stores embeddings and finds the texts closest to a question.

Common mistakes

  • Thinking semantic search always replaces keyword search (hybrid search is often better).
  • Forgetting that the quality depends on the embedding model.
  • Confusing semantic search (meaning) with full-text search (words).

Quick checklist

  • First I check whether the word names a concept, a tool, a risk, or a metric.
  • I tie it to a concrete case: I search 'how to reduce customer attrition.' Semantic search finds a doc that talks about 'fighting churn.' A keyword search misses it.
  • I keep the main trap in mind: Thinking semantic search always replaces keyword search (hybrid search is often better).

Quick questions

What is Semantic search in AI?

Semantic search finds the meaning of a question, even when the exact words change.

Where will I run into Semantic search?

You run into it in RAG, embeddings, vector databases, and internal search engines.

Which word should I read next?

Start with Embedding, Vector database, Hybrid search.

Want to keep going in order?

Head back to the full glossary, search a word, then open only the pages that deserve more than a short definition.

Open the AI glossary