Glossary
What is knowledge base?
The private collection of pages, files and snippets an AI support agent searches before every reply, not the same thing as a public help center.
A knowledge base is the private collection of content an AI support agent searches before every reply: website pages, uploaded documents, text snippets, question-and-answer pairs and pages synced from tools such as Notion or Google Drive. The agent does not read the whole collection each time. It retrieves the passages most relevant to the question and writes from those.
The word is overloaded. Helpdesks have used "knowledge base" for years to mean the public help center customers browse. In AI support the private collection and the public help center overlap, since a published article is usually fed into the private one, but they are not the same thing, and a vendor's "knowledge base" limit may count either. Ask whether a figure counts sources, pages, articles or megabytes before comparing plans.
Where the idea runs out: the agent knows only what the knowledge base says, in the words it says it. A page that answers "Do you ship to Canada?" with marketing copy produces vaguer answers than a plain FAQ sentence. Two pages that disagree let the agent pick either. Keeping one version of the truth, written as statements, matters more than adding volume, and the most useful maintenance is reading transcripts for the questions the knowledge base could not answer and writing the page each one needs.
Related terms
- Knowledge sourceOne item in a knowledge base, such as a crawled website, an uploaded file or a connected Notion workspace, and often the unit a plan limits.
- Retrieval-augmented generation (RAG)Searching your content for passages relevant to a question and handing them to the model as it writes the answer.
- CrawlAn import that starts at one address and follows a site's links to collect pages for the knowledge base, usually with a page and depth limit.
- GroundingMaking the model answer from source text supplied as it replies, not from training memory, so every claim traces back to your content.
See it in a real chat
PepoChat runs on this site. Ask the agent in the corner anything about the product.
