Text turned into numbers that capture meaning.
An embedding is a list of numbers — a vector — representing a piece of text, produced so that texts with similar meaning sit close together in that numeric space.
Search then becomes geometry: find the stored vectors nearest the question's vector. That is how retrieval finds a passage about "cancelling a plan" when the user typed "how do I quit".
Embeddings are far cheaper to compute than generation, which is why retrieval systems embed everything up front and generate only once.