Knowledge base
A knowledge base is a collection of documents an AI system treats as true and draws on. For a brand, it's the record of what's true about you.
A knowledge base is a collection of documents that an AI system draws on and treats as true. It can be a company's help centre, a folder of product sheets, or a set of web pages. When a model answers from a knowledge base, it looks up the relevant parts first and writes its answer from them.
For a brand in AI search, the term has a second use: your own record of what's true about your business, so you can check what the AI platforms say against it.
How AI systems use one
A support chatbot is the usual example. You give it your help articles. When someone asks a question, the system finds the passages that match and hands them to the model with the question. The answer is only as good as those articles. If an article is out of date, so is the answer.
Your brand's record of what's true
The AI platforms don't read your knowledge base. They read the web. But keeping one of your own is still worth it, for two reasons. It lets you check the answers: you can't say a claim is wrong until you've written down what's right. And it's the material your public pages should be saying clearly, so the platforms find it.
For example, Kestrel Roasters might keep a short document with its prices, its delivery areas, the date of its last menu change, and the certifications it holds. When an answer says "Kestrel delivers only in Scotland," they can see at once that it's wrong and which fact it breaks.
In AnswerOrbit
We call this your facts and your source of truth. You add facts in the Facts tab in Claims, and you can add material, like a page, your whole site, a file or some text, for us to draft facts from. We check every claim the answers make about you against the facts you've turned on. See Claims and facts.