What they are
An embedding model turns a piece of text into a long list of numbers. Texts with similar meaning get similar numbers. "Cheap CRM" and "affordable customer management tool" end up close together. "Cheap flights" ends up far away.
How AI search uses them
- Pages are split into passages, and each passage gets an embedding.
- The question gets an embedding too.
- The passages closest in meaning to the question are picked.
- Those passages are given to the AI to write its answer.
This is the retrieval step in RAG.
What it means for you
A passage that clearly answers one question is easy to match. A passage that mixes five topics gets a blurry embedding and matches nothing well. Clear, focused sections help.
Example
A help page has one section: "Exporting your data to CSV". Its embedding sits right next to the question "how do I export my data?". It's picked. A rambling page covering exports, billing and login in one block isn't.
Frequently asked questions
Do I need to create embeddings for my site?
No. Search engines and AI tools create their own.
Can I see my page's embedding?
You can create one with developer tools, but it's a list of numbers that isn't useful to read.
Are embeddings the same as keywords?
No. Keywords match words. Embeddings match meaning.