What Is Semantic Search?
Also called vector search
Short Answer
Semantic search finds results by meaning instead of exact words. It turns the query and the documents into embeddings and returns the documents whose vectors are closest to the query. It finds relevant answers even when they use different words, which keyword search misses.
- Taught in lesson 98
Why It Matters
Semantic search finds information the way people ask for it, not only by matching exact words. It is the core of modern document assistants and internal search tools, and getting it right is often the difference between an AI assistant that helps and one that answers from the wrong source.
How It Works
Keyword search, such as the BM25 method, is still better at exact matches like product codes, names and rare terms. That is why many production systems use hybrid search, combining both, and then rerank the top results with a stronger model.
Semantic search is the retrieval step of most retrieval-augmented generation systems, and its quality largely decides the quality of the final answer.
In production, semantic search is rarely used alone. Combining it with keyword search (hybrid search) catches exact terms such as codes and names, filters narrow results by date, type or permission, and a reranker reorders the top results more precisely. Quality is measured with test queries whose correct results are known, so changes can be compared.
Semantic Search in Practice
Everyday Example
Searching an online shop for "shoes for running in the rain" returns waterproof trail shoes even if no product page contains that exact phrase.
Industry Example: Pharma
A scientist asks "studies where the drug lowered blood pressure in older adults" and semantic search finds papers written in clinical terms such as reduced systolic pressure in elderly patients.
Where You Meet It
- Search boxes in modern apps and help centres.
- Retrieval in RAG assistants.
- Finding similar past cases or tickets.
- Internal knowledge bases.
Key Ideas Around It
- Keyword search
- Matching the exact words of the query.
- Hybrid search
- Combining keyword and semantic search.
- Reranking
- Reordering top results with a more precise model.
- Recall
- How many of the relevant results the search finds.
A Common Mix-Up
Semantic search is not always better than keyword search. For exact identifiers, product codes or rare names, keyword search is often more reliable, which is why hybrid search is common.
Check YourselfWhy might semantic search fail on a query for a product code like "AX-4471"?
Codes carry little meaning for an embedding model; keyword search matches them exactly, so hybrid search works better.
Learn It on ai2hired
S9 · #98Embeddings
Coming in Season 9, Embeddings, RAG and Search. Every ai2hired lesson is free on YouTube.
Questions People Ask
Is semantic search better than keyword search?
It is better for meaning and worse for exact identifiers. Combining both in hybrid search usually works best.
What do I need to build semantic search?
An embedding model, a place to store vectors such as a vector database, and code to embed queries and return the nearest results.
Related Terms
Learn AI From Zero to Hired
Free lessons on YouTube, each one built from what real AI job ads ask for.