In plain language
It lets a search find “refund request” when the customer wrote “I want my money back.”
Example
Help articles are converted to embeddings and stored, then the closest matches are retrieved for each new question.
Related terms
Retrieval-Augmented Generation (RAG) · Large Language Model (LLM) · Classification · Token (AI)
Common misconception
Embeddings from different models are not interchangeable; changing models means re-creating stored embeddings.
Implementation guidance
Use one embedding model consistently, re-embed content when it changes, store source links with each embedding, and evaluate search results on real queries.