Product

Knowledge Base RAG Chat Wisbot Helpdesk

Solutions

Customer service AI website assistant Internal knowledge base Cities & public sector All solutions
Pricing Contact Log in Get a demo
Semantic search

Semantic Search That Understands Meaning

Our search finds content by meaning, not just by keywords. A question worded nothing like your documents still finds the right answer.

Get a demo
On this page
  1. Finding answers by meaning
  2. Proven technology underneath
  3. The engine behind every answer
  4. Frequently asked questions

Finding answers by meaning

Customers never phrase a question the way your documentation phrases the answer. Keyword search fails here: if the words don't match, nothing is found.

Semantic search solves this. It compares meaning, not spelling. A visitor asks: "Can I get my money back?" And finds your text: "Refunds are processed within 5–7 business days." Not a single word in common — and still exactly the right answer.

Proven technology underneath

How does that work? Every piece of your content is converted into a so-called vector — a mathematical fingerprint of its meaning. We store these in PostgreSQL with the pgvector extension, a proven, widely used database technology.

For you this means: no extra systems to run, fast answers even with large collections, and all data hosted in Germany.

The engine behind every answer

Semantic search does the groundwork for the whole platform. When the chat or your website chatbot answers a question, this search finds the right content first. It works best when the content is well prepared — that's the job of smart chunking. And everything you add — Q&A pairs, web pages, documents — is searched together.

Frequently asked questions

How is semantic search different from keyword search?
Keyword search needs the same words in the question and the answer. Semantic search compares meaning. A customer asks in their own words — and still finds your documentation, written in yours.
What technology is behind it?
Your content is converted into vectors by an AI embedding model and stored in PostgreSQL with pgvector. In plain terms: the system measures how close two texts are in meaning. That stays fast, even with large collections.
Does it work across all my content?
Yes. Q&A pairs, imported web pages and uploaded documents are all searched together. The best answer wins — no matter where it is stored.

Ready to support
smarter?

Book a personalised demo — we'll walk you through it.