What is a broken link?
A broken link is a hyperlink whose intended destination cannot be reached or no longer supplies the promised resource. A missing-page response is one possibility; a failed request or an unrelated replacement can also interrupt the task. Diagnose the link as the reader follows it, including any redirected destination.
Keep the referring page and the linked address separate. A valid page can contain a mistyped link, while a correct link can lead to a resource that was accidentally removed. The repair depends on which part is wrong and who controls it.
Prioritize links that block a reader's task
Start with the links people need to use: navigation, booking instructions, download references and other paths necessary to complete a task. A bad address in a shared template deserves attention because it can appear in many places. Record the affected pages instead of assuming that every reported error has the same consequence.
Use available visits, requests and reader reports to understand priority, while keeping missing data unknown. Backlinks can identify an old address that people still follow, but their presence does not make an unrelated destination useful. Decide what that visitor expected before choosing a replacement.
Trace the link before choosing a repair
Record where the link appears, what its text promises and the exact address it requests. Then inspect both the response and the final visible resource. A crawler finding is a starting observation, rather than a complete explanation of why the link failed.
Check what changed
Look for a typo, a renamed page, a removed file or a routing change. If the destination is unavailable only in a particular context, check access requirements and service conditions before declaring it permanently gone. Compare the current address with the intended publication or release record.
Collect the link and its observed response
For a reported problem, open the referring page and follow the actual link. A crawler can help collect repeated failures across a site; inspect its results before deciding which addresses need repair. Keep the requested URL, any redirect destinations, response status, visible content and observation date together.
Google's HTTP-status documentation explains that a server generates status codes in response to requests. Read the code as evidence about that request, and inspect the page itself for the information the link promised.
Choose restoration, a link update or a relevant move
If the intended resource should still exist, investigate restoring it or fixing the serving problem. If the referring link contains an incorrect address, update it where you control the source. If the resource has genuinely moved, map the old address to its appropriate new home and update controlled links to that destination.
Google's crawl-budget guidance recommends returning 404 or 410 for permanently removed pages. When there is no useful replacement, keep that truthful missing-page response and correct references that still promise the retired content. Do not redirect to a homepage simply because the old URL has backlinks.
A fictional timetable replacement
Imagine a fictional community center replacing a downloadable timetable. Its current activities page links to the new document, but an older article still links to the removed file. An editor follows both links and confirms which timetable is intended for readers today.
In this example, the editor updates the old article's reference. Before adding a redirect for old bookmarks, they check whether the new document actually replaces the old one or covers a different activity. A document with a different purpose needs a different decision, even if it loads successfully.
Separate response codes from destination content
For Google Search, a successful 2xx response does not guarantee indexing. Google's HTTP-status guide says an empty page or error message can instead be reported as a soft 404. A successful request is therefore insufficient to establish that the link delivers useful content.
The same guide groups 4xx responses other than 429 together for Google's handling: Search does not index those URLs and removes previously indexed ones after processing the unavailable response. It supplies no guaranteed faster-removal deadline for choosing 410 over 404. A list of such responses is not a calculation of sitewide ranking loss.
Use the observed condition to choose the next check. These are investigation prompts, not proof of a particular cause.
| Observed condition | Next check | Repair decision |
|---|---|---|
| 404 or 410 response | Should this exact address still provide content, or was it retired? | Restore an unintended loss, map a genuine move or update the referring link. |
| 5xx or 429 response | Check service availability and relevant request logs. | Investigate the serving condition before treating the resource as deleted. |
| No completed response | Retest the hostname, address and connection context. | Resolve the failed request before selecting a permanent content change. |
| 200 response with an error or empty page | Inspect the rendered content and required resources. | Restore intended content, or return an appropriate missing-page response if none belongs there. |
| Redirect reaches the wrong resource or loops | Follow the route and compare the final page with the link's promise. | Correct the mapping or conflicting rules, then update controlled references. |
Verify both ends after release
Keep a repair record with the referring page, original destination, chosen action and reason. Test the released change from the referring page, rather than checking only an edited template or preview. Revisit any addresses left unresolved and record what still needs an owner.
- Activate the link as a reader would, including any fragment that points to a section.
- Inspect the requested response and any redirect route.
- Check the final content against the original task and link text.
- Confirm that shared navigation and other controlled references use the intended address.
Schedule checks around meaningful changes
Review important paths after changes to page names, templates, files or hosting rules. Choose ongoing checks according to how often those resources change and how people use them. A release affecting shared navigation needs a broader check than an isolated spelling correction.
Google's advanced crawl-budget guide is intended primarily for very large or frequently updated sites, with additional scope for sites reporting many discovered but unindexed URLs. Do not assume every broken link represents an urgent crawl-capacity problem. Start with the failed reader journey and the actual response evidence.
Repair the reference without hiding the problem
For an outbound source link, check whether the surrounding claim is still supported. A different live page is not automatically an equivalent reference. Replace the citation with appropriate evidence, revise the claim or remove unsupported material, rather than changing the URL while leaving a misleading statement intact.
For addresses you control, a 301 redirect concerns a permanent move, while a 404 error concerns the missing response. Keep those concepts separate from broken link building, the outreach task of proposing replacement resources to other publishers. Repairing your own reader's path does not require an outreach campaign.
Primary references
- Google Crawling Infrastructure: HTTP status codes and crawler handling.
- Google Crawling Infrastructure: crawl-budget scope and retired-page guidance.
Explore this topic: Broken Link Building: A Step-by-Step Outreach Guide.