Every Confluence site collects pages that are no longer the right answer: the same runbook written twice, a guide that a newer one replaced, a page whose owner left years ago. People find them in search and follow them. AI assistants such as Rovo answer from the same content, and they cannot tell that one of two nearly identical pages was abandoned.
Whether or not you use Rovo yet, a clean-up pass pays off. Here is how to do one with what Confluence Cloud already gives you.
1. Know what you are looking for
Three kinds of pages cause most of the wrong answers:
- Duplicates - the same content written twice, often by two teams that did not know about each other, or a "Copy of ..." that became the real page while the original stayed.
- Superseded pages - an older version of a guide, policy or process that a newer page replaced, with nothing on the old page saying so.
- Pages nobody owns - content whose owner has left. It may still be right, but there is nobody to ask, and nobody will update it when it stops being right.
Old is not the same as wrong. A five-year-old policy can be perfectly current. The goal is not to delete everything old - it is to make sure every page that answers a question is the one you want quoted.
2. Find stale pages with search and CQL
The quick way: in Confluence search, use the Last updated filter to list pages not updated in the past year, and narrow it to one space at a time.
For more control, CQL. While signed in, open this in your browser (replace your-site and the space key):
https://your-site.atlassian.net/wiki/rest/api/search?cql=type=page AND space=HR AND lastmodified < now("-2y")&limit=100
You get the matching pages as JSON. Useful variations:
lastmodified < now("-1y") for a stricter cut-offtitle ~ "copy of" to find copied pages that were never renamedtitle ~ "old" OR title ~ "deprecated" OR title ~ "v1" for pages someone already suspected were outdated
Start with the spaces people ask about most - HR, IT, engineering runbooks - rather than the whole site. If your plan includes analytics, the most-viewed pages in those spaces are the ones that matter first.
3. Find duplicates
This is the hard part, because duplicates rarely share a title. Some ways in:
- Sort a space's page tree by title and look for near-matches ("Onboarding Guide", "New Starter Onboarding", "Onboarding 2024").
- Search for a distinctive sentence from a key page in quotes. If it shows up on two pages, one of them is probably a copy.
- Ask the space owners. They usually know which pages they forked and forgot.
Be careful with pages that look alike on purpose: a runbook per region, an API guide per version, holiday calendars per country. Those are parallel, not duplicate, and both should stay.
4. Find pages whose owner has left
Every page shows its owner near the title ("Owned by ..."). That field follows reassignment, so it is a better signal than the creator. For a space with a few hundred pages, compare the owners against your current staff list (your HR or identity team can export one). For pages whose owner has gone, reassign them to the team that now owns the topic before deciding anything else.
5. Decide what to do with each page
For every page you find, one of four things:
- Keep - still correct. Optionally reassign the owner so someone is responsible.
- Redirect - add a short note at the top: "Replaced by [newer page]" and link it. Useful when people still have the old page bookmarked.
- Merge - move anything unique into the page you are keeping, then archive the other.
- Archive - take it out of everyday use without losing it.
A few things worth knowing about archiving in Confluence Cloud:
- Archived pages can be restored at any time, so archiving is a safe default when you are unsure.
- Archived pages are hidden from Confluence search by default.
- Archiving emails the page's watchers. Tell the space owners before a big batch, so nobody is surprised by a flood of notifications.
- Archiving needs Confluence Standard or above; the Free plan cannot archive. Bulk archiving of many pages at once is Premium and Enterprise.
6. Make it a habit
A one-off clean-up drifts back within a year. A light routine keeps it in shape:
- A quarterly look at the stale-page search for your top spaces.
- When someone leaves, reassign their pages as part of offboarding.
- When a page is replaced, archive or redirect the old one in the same change.
A faster way, if you have a large site
Disclosure: I'm the developer of Supersede at AIGA Tech, a Marketplace app built for exactly this.
The manual steps above work well for a few spaces. On a site with thousands of pages, the duplicate search is where it breaks down, which is why I built Supersede. It runs on Atlassian's own platform and:
- fingerprints every page to find overlapping content, even when titles differ,
- uses Atlassian-hosted AI to judge each pair - redundant, or deliberately parallel like the regional runbooks above - and says why in one sentence,
- lists stale pages and pages whose owner no longer has access,
- lets an admin approve, ignore or archive in bulk (archiving on Standard and above), with a report and CSV export.
The AI only advises; an admin decides every action, and page text is never stored.
Supersede on the Atlassian Marketplace: https://marketplace.atlassian.com/apps/3108262762/supersede
Happy to answer questions about any of the steps above, with or without the app. How do you keep your own Confluence site tidy - a regular clean-up, or as you go?
