Vector database
A vector database stores text as lists of numbers that capture meaning, so an AI system can find passages that mean the same as a question.
A vector database is a store for embeddings: lists of numbers that capture what a piece of text means. When an AI system needs passages that answer a question, it turns the question into numbers too, and asks the database for the passages whose numbers are closest.
You won't run one to get named in AI answers. But many AI systems use one to decide which passages to read, so knowing how it works explains why clear, focused writing gets found.
How it works
Text is first split into short pieces, often a paragraph or a few sentences each. Each piece is turned into an embedding and saved. Pieces that mean similar things end up with similar numbers, even if they use different words.
When a question comes in, the system turns it into an embedding and looks for the nearest pieces: the ones closest in meaning, whatever words they use.
An example
For example, a shopping assistant has stored passages from hundreds of coffee pages. Someone asks: "Which beans are good for a French press?" The nearest passages might include a paragraph from Kestrel Roasters' brew guide headed "Best beans for plunger coffee," even though it never says "French press," because the meaning is close.
A paragraph that covers French presses, filter machines and espresso all at once is a muddier match for each of them, so it's less likely to come out on top for any one.
What it means for your writing
- Give each section one clear subject, so it matches the questions about that subject
- Use the words your customers use, alongside your own terms
- Make each paragraph make sense on its own, since it may be found and read without the rest of the page