I’ve been noticing this in our own work. We use the Atlassian MCP quite a bit with AI agents, and agents end up creating pages, notes and plans as part of their work.
It’s useful, but over time you get multiple versions of the same thing, duplicate pages, slightly different instructions, and old questions that nobody has gone back to answer. Across multiple teams, it gets surprisingly hard to know what’s actually still true.
I ran into this often enough that I ended up building Docgrity, a small Confluence app that checks your documentation for:
Duplicates — pages that are essentially saying the same thing
Contradictions — different pages making conflicting claims
Open questions — unanswered questions and stale TODOs
The bit I care about most is that it doesn't just say “these pages might conflict.” It shows the actual text behind the finding, so you can decide for yourself whether it's a real problem. It can also surface potential owners based on page history.
It’s still early, and I’m trying to figure out whether this is a problem other teams are seeing too.
If you’re using AI agents with Confluence, how are you dealing with the documentation they leave behind? Are duplicates, conflicting versions and stale information becoming a problem?
I’d really like feedback on whether this is useful, what I’m missing, and especially where the false positives are.
I’ve put the early version of Docgrity here.
