How to Find Broken Links in Confluence Cloud

The example: two suspect links

The IT support runbook in the Support Handbook space has two links picked for this review.

1 2 Enlarge Enlarge screenshot

On 2 October 2026 both destinations returned HTTP 404. Proving that is only the first half of the job. The review is finished when the right replacement is found, the source page is updated and the reader’s path works again.

A page that opens is not automatically the right page either. A link that lands on a generic homepage or on outdated instructions fails the reader as surely as a 404.

1. Choose the pages that matter

“Check the whole knowledge base” is rarely a real task. Start with a set of pages people actually use:

• onboarding pages new colleagues follow;
• support articles agents send to customers;
• runbooks used during incidents;
• a space affected by a domain or documentation move.

Write the page list down before you start. “No broken links found” only means something when the reader knows which pages were checked.

On each page, include the links people click: text links, buttons, Smart Links, attachment links and links to a heading. To check page for broken links properly, look past the obvious blue text and confirm that each destination is still the intended one.

Check the Undefined list separately

Confluence has one native list related to links: Space settings→Content→Undefined. It sits in the space settings, so you need access to them.

1 2 3 Enlarge Enlarge screenshot

The list covers links to Confluence pages that do not exist yet. It does not test external addresses, so an empty list does not mean the space has no broken links.

Check: you have a written page list, and you have looked at Undefined if unfinished placeholder pages are in scope.

2. Open each link as the reader

Open the published page and follow each link with the access your readers normally have. An administrator can often open a page that an agent cannot, so a link that works for you may still fail for them.

For the IT support runbook, VPN setup leads to a 404 on example.com, and Confluence API notes leads to this:

1 Enlarge Enlarge screenshot

If a link points to a heading, confirm that it lands on that section. If it points to an attachment, confirm that the reader can open the file.

Classify the result before you fix anything

A 404 is strong evidence that the page is gone from that address. A sign-in page is not, and a timeout is weaker still. This matters every time you check for broken links: replacing a correct restricted URL because it asked for a login makes the documentation worse.

Check: for every link you tested, you can say what happened in words, not only “broken”.

3. Record the source, result and next action

Use one row per link on a source page. If the same address appears on three pages, write three rows: the pages can have different owners and get fixed on different days.

Write a concrete next action. “Find the current VPN instructions” moves the work forward; “broken” does not. A useful register also has the destination URL, who you checked as, the owner, the replacement URL and the recheck date and result. When every broken link check row shows its source page and date, the next broken links check can start from the last one instead of from zero.

Check: every failed or uncertain link has an owner and a next action, not just a status.

4. Fix the link on the source page

Find the replacement first. Open it and make sure it answers the same question as the old page did. Then change the link:

1. 1

   Step 1: Open the link dialog

   Open the page, select Edit, click the link text and choose Edit link in the small toolbar that appears.

2. 2

   Step 2: Replace the address, then publish

   1 2 3 4 Enlarge Enlarge screenshot

   Paste the new address, review the display text, select Save in the dialog and then Update on the page. Until you select Update, readers still see the old link.

For a link to a section of a Confluence page, do not guess the anchor. Atlassian’s method: hover to the right of the heading until the link icon appears and select it to copy the heading link. Atlassian: Insert links and anchors

Check: the published page shows the new link, and it leads to the intended content.

5. Recheck before you close the row

After a fix, walk the reader’s path again: open the published page, follow the link, confirm access, confirm the content, and for a heading or attachment link confirm the exact target. Then write the date and result in the register. Do not close a row just because the editor accepted the URL.

If you changed four links, that is four rechecks. If the same address was fixed on four pages, open it from all four.

Every branch that changes a link ends the same way: edit the source, select Update, open it again from the published page, write down the result. That is the whole answer to how to check broken links by hand: define the pages, test, classify, fix and recheck.

Before you close the review

• The list of checked pages is written down.
• Every changed link has a verified replacement.
• 404/410 results are kept apart from access problems and timeouts.
• Restricted links were checked with the reader’s access or by their owner.
• Display text still describes the destination; heading links land on the right section.
• Changed pages were updated, not only saved in the link dialog.
• Open uncertain results have an owner and a next action.

Why a website scanner may not see your Confluence pages

An online broken link checker can only scan the pages it can open. A crawler that is not logged in cannot read Confluence pages that need a login, so a broken link checker online can return a clean report without ever reading the private pages your team uses.

The same applies to a website broken link checker, a website link checker or any link checker online. Before you rely on one to check website for broken links, ask which identity opens the source pages. Requests like “check site for broken links” or “check my site for broken links” assume the crawler can find and read the site; for a private Confluence space that is often false.

There is a second gap. A general broken link checker tool or link checker tool may report the source URL, but it does not know who owns a Confluence page. Check that each result leads back to the page that contains the link and to the person who should fix it.

When a scanner starts to pay off

The manual method is fine for a few pages now and then. The cost grows with the page count: 5 pages × 10 links is up to 50 opens; 20 pages × 15 links is up to 300; every fixed link adds a recheck.

At that scale a broken link finder or dead link finder saves the discovery work, but someone still reads the results. A 404 broken link checker is useful for missing pages, as long as it does not treat sign-in pages and timeouts as the same thing. A dead link checker should show the evidence behind “dead”. A broken link scanner should say whether the scan finished or skipped content. And a link validator can confirm that a URL responds, not that it holds the right instructions.

Whatever broken links checker you consider, including the tools a search for “broken link checker confluence” returns, check four things: which Confluence content and link types it reads; how it handles restricted pages; whether each result points back to its source page; and whether you can tell a fixed link from an old result.

Common questions

If you check broken links across many spaces, Broken Link Checker for Confluence scans selected spaces and keeps each result tied to its source page and owner. It is free for up to 10 Confluence users.