Auditing Confluence pags created via Rovo or other AI assistants via Rovo MCP

[Mini Channanathil]

Hi,

Since we have AI tools in place and also Rovo enabled within our organisation, from a Security and Governance perspective, how do we ensure that Users actually review the content that is published.

For example, for a use case where Users generate content in Claude or ChatGPT and then copy it over to a newly created page on Confluence, the page ID is published on the audit logs within Atlassian. But if a User creates a Confluence page directly from Claude/ChatGPT, the audit log only mentions that mcp tool was invoked, but there are no details about the page created.

Hence, from an admin perspective, it is difficult to identify AI generated content. Our goal is to ensure that users will review the content they generate and hence we want to put guardrails in place to ensure it is done.

Has anyone any ideas on how to achieve this?

Best regards,

Mini

[Peter_DevSamurai]

Hi Mini, 

The Confluence audit log does capture the page creation, you can combine that with automation rules for labeling/notifications and page approvals for a solid review workflow.

1) Cross-reference wth the Confluence audit log
Even when a page is created via MCP, the Confluence-level audit log still records a page_created event with the user who triggered it, the page title, space, and timestamp. So the data is there, it's just in a different log than the org-level one.

2) Use Confluence Automation to flag new pages

Set up an automation rule that triggers on "Page created" and automatically:

• Adds a label like "needs-review"

• Sends a notification to a reviewer or admin

• Posts a comment prompting the author to confirm they've reviewed the content

This creates an enforceable review step regardless of how the page was created.

3) For Premium/Enterprise plan, use Page Approvals
If you're on Premium or Enterprise, enable Page Approvals on key spaces. This prevents content from being considered "approved" until a designated reviewer signs off, which is  perfect for AI-generated content governance.

I hope this can assist you with your inquiries.

Bestest regards,

Peter Tran