vector database
Also called: vector db, vector store, vector search database
A vector database stores data as embeddings (high-dimensional numeric vectors) and retrieves records by semantic similarity rather than exact keyword match. It powers the retrieval step in RAG pipelines that ground AI assistants like ChatGPT and Perplexity in outside documents.
How it works
Instead of indexing words, a vector database indexes meaning. An embedding model turns each chunk of text (or image or audio) into a list of numbers that captures its semantic properties, so passages about “running shoes” and “trainers for jogging” land near each other in vector space even with no shared words. At query time the database runs approximate nearest neighbor (ANN) search over structures like HNSW, then ranks candidates by a distance metric such as cosine similarity or dot product. ANN trades a little accuracy for speed: the more exact the result, the slower the query.
This is the engine behind retrieval-augmented generation (RAG). In a RAG pipeline your content is chunked, embedded, and stored as a “knowledge library” the model can read. The user’s question is embedded too, matched against that store, and the closest passages are pasted into the prompt before the LLM writes its answer. Grounding generation in retrieved passages is how assistants cut hallucinations and point to current sources instead of relying only on training data.
A rough flow:
page text -> chunks -> embeddings -> vector store
user query -> embedding -> ANN match -> top passages -> LLM answerCommon stores include Pinecone, Weaviate, Milvus, Chroma, and Postgres with the pgvector extension. For SEO the takeaway is simple: getting pulled into an AI answer is a retrieval problem before it is a ranking problem, and retrieval runs on vectors.
How it affects your traffic
When ChatGPT, Perplexity, or Google's AI Overviews answer a question in your market, a retrieval step decides which pages get pulled into the response, and that step runs on a vector database. If your content is chunked into clear, self-contained passages that embed close to how buyers actually phrase questions, you get retrieved and cited. If the same facts are buried in long, unfocused blocks, you can get skipped even when you rank well in classic search. Structuring pages so they survive that embedding-and-retrieval step is the core of our AI SEO work.
Get AI SEO that moves the needle
We turn terms like this into ranked pages and qualified pipeline. Start with a free Initial SEO Strategy.