Skip to main content
All CollectionsKnowledge Base
Verifying documents in the knowledge base
Verifying documents in the knowledge base

Ensure the accuracy and relevance of your knowledge base documents

Jake Rosenthal avatar
Written by Jake Rosenthal
Updated over a week ago

How to verify documents in the knowledge base

(*Document Verification is currently only available for Enterprise members.)

This feature is useful for verifying the status and relevance of documents in your knowledge base, helping you maintain its accuracy and improve AI output over time. By regularly cleaning up your knowledge base, you can ensure that the AI doesn't base its answers on outdated references or irrelevant information.

  1. Access document verification in chats: While in a chat, click the "View Document Verification Status" badge icon to display the status of each reference, along with an option to update it.

  2. Update document status: Click on the dropdown menu next to a document to mark it as verified, request verification, or flag it as outdated. If a status has already been set, you can clear it using the same dropdown.


    โ€‹
    โ€‹*You can optionally include a comment when requesting verification or flagging as outdated.

  3. Verify documents in the knowledge base: When viewing a document directly in the knowledge base, you'll find the same options to mark it as verified, request verification, or flag it as outdated.

  4. Review verification requests and outdated items (admin): As an administrator, you can review items that were previously verified or marked as outdated. You can also quickly approve or reject documents that have been requested for verification.

  5. Exclude outdated items from AI outputs: By flagging an item as outdated, it will no longer be used by the AI when generating outputs, ensuring that the AI relies only on up-to-date and relevant information.

By regularly verifying and updating the status of documents in your knowledge base, you can maintain its quality and improve the accuracy of AI-generated outputs, ultimately enhancing the overall user experience and the effectiveness of your Cassidy implementation.

Did this answer your question?