Internal knowledge enables you to find answers faster, reduce context-switching between tools, and make more informed decisions by bringing the most relevant internal information from connectors, or connected apps, into ChatGPT’s response. Please note that connectors like Google Drive are provided by third parties and also subject to their terms of use, and your workspace admin can choose which connectors to enable or disable for internal knowledge for your workspace.

ChatGPT can infer from your prompt when to use internal sources. You can also select “Internal knowledge” in the message composer to direct ChatGPT to pull information from your organization’s connectors, such as Google Drive.

Currently, users can only connect internal knowledge with Google Drive. We support a range of file types from Google Drive, including:

Please note that images and charts embedded within these files aren’t supported at this time.

Google Sheets and Excel workbooks are partially supported. The model can search and read these spreadsheets for basic queries, but does not support advanced data analysis.

Drive connections must be made to a Google Workspace’s Google Drive, not personal Google Drive accounts. The number of connections per workspace is capped. If you choose to disconnect the Google Drive connector, it will immediately stop syncing and all indexed content is removed from that source.

ChatGPT Workspace Admins can configure internal knowledge with one of the following methods:

ChatGPT Workspace Admins have three options for managing shared drives:

We are doing a public beta release of this product as we’re still testing. We are gradually rolling this out to Team workspaces over the next few weeks. Team admins will get notified in product when they are selected for this early release and they need to configure the connector before end users can see this in product.

Permissions are fully respected and kept continuously up to date. Your employees will not be able to discover content via ChatGPT which they cannot access in Google Drive.

This means each employee may receive different responses for the same prompt.

ChatGPT keeps continuously up to date on both changes to files/channels as well as users' access to them. For Google Drive, the exact mechanism of this differs depending on how you connect — either we directly record whether a user has access to a file or we sync the permissions for the file itself and associated directory information in order to resolve group membership.

Internal knowledge requires a one time initial syncing of files; this will take some time as we index all data from this source, enabling real time retrieval in context of your chatGPT querie while respecting all file permissions. After this initial sync files/permissions are automatically updated on a regular basis (within minutes) to reflect recent changes—no action is required to keep your content up to date.

Initial syncing can vary depending on how many files you have access to, but happens in a few stages once you connect a source:

Internal knowledge is initially designed to work best for Q&A and search related queries. The most relevant data is sent to the model based on query intent, limiting performance in scenarios requiring aggregation from numerous sources or very complex queries, such as financial data aggregations.

Currently, connections made for internal knowledge cannot be updated.

Currently, only the Windows app has parity with the full experience on ChatGPT.com. We are working toward bringing the full internal knowledge experience to the rest of our apps - please stay tuned!

Currently, you cannot reference a document or conversation by linking the URL directly in ChatGPT.

We’re working on connecting ChatGPT to more tools teams rely on every day—from docs and collaboration to data, CRM, and more.