Skip to main content

AI glossary Models & Architecture

What is an embedding in AI?

← All glossary terms
In short

A list of numbers (a vector) that represents the meaning of a piece of text, an image or audio, so that similar meanings sit close together.

Explained

An embedding is a dense numerical vector produced by a model to represent the meaning of a word, sentence, document, image or sound. The model is trained so that items with similar meaning end up close to each other in that vector space, even when they share no words. Comparing vectors, usually with cosine similarity, lets software find related content, group it or detect duplicates. Embeddings are the basis of semantic search and of the retrieval step in RAG systems.

How embeddings work

An embedding model reads an input and outputs a fixed-length vector, often with hundreds or thousands of dimensions. Each dimension has no meaning on its own; what matters is the position of the vector relative to others.

To search, you embed your documents once and store the vectors, usually in a vector database. At query time you embed the question with the same model and look for the nearest vectors. The closest documents are the ones most similar in meaning.

What they are used for

  • Semantic search over documents, tickets or products.
  • The retrieval step of RAG, which feeds relevant passages to an LLM.
  • Recommendations: items close to what a user has already seen.
  • Clustering and topic discovery in large text collections.
  • Detecting duplicates and near-duplicates.
  • Classification with few labelled examples.

Choices that affect quality

  • The embedding model: general or domain-specific, multilingual or not.
  • How you split documents into chunks before embedding them.
  • Dimensions and storage cost, which grow with the size of the collection.
  • Keeping the same model for documents and queries: vectors from different models are not comparable.
  • Re-embedding the collection when you change model.

Frequently asked questions

What is an embedding in simple terms?

It is a way to turn text, images or sound into a list of numbers that captures what they mean. Two sentences that say the same thing in different words get similar lists of numbers, so a computer can tell they are related.

What is the difference between an embedding and a vector?

A vector is just an ordered list of numbers. An embedding is a vector produced by a trained model so that its position encodes meaning. Every embedding is a vector, but not every vector is an embedding.

Are embeddings the same as an LLM?

No. An LLM generates text; an embedding model turns input into a vector and generates nothing. They often work together: embeddings find the relevant documents and the LLM writes the answer from them.

Where are embeddings stored?

Usually in a vector database or in a search engine or database with vector support. These systems index the vectors so the nearest ones can be found quickly, and they let you filter by metadata such as date, owner or permissions.

Do embeddings work across languages?

Multilingual embedding models place sentences with the same meaning close together even when they are in different languages. That lets a question in Spanish find a document written in English. Quality varies by model and language, so it is worth testing on your own content.

Decision guides that use this term

How to put it to work

Where it shows up at Thinkia

Further reading

Sources

Updated: 2026-10-21

Want to see how this applies to your organisation?

Talk to an AI Expert