First, separate pages that should be gone from pages that should not
The crawling and indexing guide states how a not-found or gone HTTP Status CodeThe numeric code a web server returns with every response, saying whether the request succeeded, was redirected or failed and why. is treated: the URL drops from the index once the response is re-confirmed. Its controls table gives the right way to remove a page permanently: a not-found or gone response plus removal of internal links, not a redirect to the home page, which is treated as a Soft 404A page that returns a success status code but shows an error, empty or not-found message.. The expired pages guide draws the conclusion for the report: not-found entries for pages removed on purpose are expected, while not-found entries for live pages, and links still pointing at removed ones, are the ones to investigate.
Google finds URLs by following links and reading sitemaps, and it keeps a list of the URLs it knows. So every address in the not-found list came from somewhere: a page that once existed, a link on the site, a sitemap entry, a link on another site, or a link written wrongly. Export the examples from the Page Indexing report and give each one two answers. Should this address exist? And where is something still pointing at it?
Then rank by value, using the Pareto principle: a small share of the dead URLs usually carries most of what is at stake. For each, note whether internal links point at it (from a crawl), whether other sites link to it (from a backlink tool and the server logs), and whether it had clicks before it died (from the performance report). A dead URL with none of the three can wait; one with all three is urgent.
| Group | What it looks like | What to do |
|---|---|---|
| Expected | Removed on purpose; nothing links to it; no successor | Leave it answering not found or gone |
| Leaking | Removed on purpose, but the site's links or sitemap still name it | Remove the references at source |
| Losing | Moved or deleted, and other sites or past searchers still arrive | Redirect to a genuine equivalent |
| Mistaken | A page that should be live answers not found | Restore it |
| Noise | An address nobody created on purpose | Stop the source; leave the response as it is |
Follow the Search Console error triage SOP for the rest of the procedure: reproduce a few examples, compare the first detected date with the release history, assign, validate and log.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Retired items, ended events or deleted posts with no links and no successor | Content log; inlinks; links from other sites | Pages removed on purpose that should stay gone | Nothing; record them as expected |
| Removed URLs still found in a crawl or the sitemap | Crawl inlinks to the URL; sitemap entries | Internal links, navigation or the sitemap still naming removed URLs | Remove or repoint the links and drop them from the sitemap |
| A page's content lives on at a new address, but the old one is dead | Compare the dead URL with live pages; redirect rules | Addresses changed without a redirect | Add a permanent redirect to the new address |
| Dead URLs receive visits or links from other sites | Backlink tool; referrers in logs and analytics | Links from other sites to moved, deleted or mistyped addresses | Redirect to the closest equivalent, or ask for the link to be corrected |
| A page that should be live returns not found | Request it; release log against first detected date | Live pages answering not found by mistake after a release | Restore the page or the route and request indexing |
| Addresses nobody created on purpose: odd paths, broken parameters, doubled folders | Find the link or script that generates them | Addresses invented by scripts, parameters or malformed links | Fix the generator; let the addresses keep answering not found |
Fix, verify and prevent each cause
Pages removed on purpose that should stay gone
Retired products, ended events, filled vacancies and deleted thin posts belong here. The expired pages guide sets the decision for each kind of page with an end date, and removal with a not-found or gone response is one of its legitimate options.
Fix: nothing, beyond confirming that each URL has no successor worth redirecting to and no links still pointing at it. Do not redirect them to the home page; a redirect is only for a genuine equivalent.
Verify: each sampled URL returns not found or gone, and a crawl finds no links to it.
Prevent: keep a list of intended removals with dates, so the next person to read the report can mark these entries as expected in minutes rather than re-investigating them. The SOP for removing a web page adds each address to that list as part of retiring it, after clearing its links, sitemap entry and channels.
Internal links, navigation or the sitemap still naming removed URLs
A removed page is one thing; a site that still links to it is another. Every Internal LinkA link from one page to another page on the same website. to a dead URL sends visitors and crawlers to a dead end, and a sitemap that lists it asks Google to fetch a page that does not exist. The internal linking guide names links to removed pages among the faults every crawl finds and says they are fixed at the source.
Fix: repoint or remove the links in templates, navigation, footers and body copy, and regenerate the sitemap so it lists only live pages.
Verify: a fresh crawl finds no inlinks to the removed URLs and the sitemap no longer contains them.
Prevent: make link and sitemap updates part of removing a page, the same day, as the expired pages guide asks. A not-found event in analytics, built as the tutorial on tracking site errors and failed actions describes, shows which links people still follow to dead pages.
Addresses changed without a redirect
A renamed product, a reorganised section or a new URL pattern leaves the old address dead unless someone writes the redirect. Google's redirects documentation, as the permanent redirect entry summarises it, describes a permanent redirect as the way to pass an old URL's role to its new home.
Fix: add a one-hop permanent redirectA permanent server redirect telling browsers and search engines a URL has moved for good. from each old address to the page that now serves the same purpose.
Verify: each old address returns a single permanent redirect to the matching page, which the SEO page checker confirms for one URL at a time.
Prevent: change addresses only through the SOP for changing a page's URL, and record every change in a Redirect MapThe table of every old URL and its new destination, one row per address, from which a site migration or URL change is built and then checked.. For a whole section or site, the site migrations guide builds the map from every source of old URLs; if traffic has already fallen after a move, the page on a traffic drop after migration takes over.
Links from other sites to moved, deleted or mistyped addresses
Other sites keep linking to the address they found, sometimes with a typing mistake. The link building methods guide finds broken inbound links with backlink tools and server logs, and notes the fix is easy to grant because the author already chose to reference the site; the link reclamation tutorial turns that into a step-by-step routine.
Fix: redirect each linked dead URL to its closest genuine equivalent; where the link is mistyped, redirect the typo too if it is common, or ask the linking site to correct it. Where no equivalent exists, let it stay gone rather than redirecting to something unrelated.
Verify: request each linked address and confirm it lands on a relevant page in one hop.
Prevent: before removing or moving any page, check whether other sites link to it, and include those addresses in the redirect map. Once redirected, they move to the list the page on page with redirect in Search Console explains.
Live pages answering not found by mistake after a release
The worst entries are pages that should be live, knocked out by a changed route, a renamed template, a replaced redirect rules file or a content record unpublished by accident. The SEO checks for staging and releases list a replaced rules file among the ways a release breaks search, and ask for the Page Indexing report to be read for new not-found entries in the week after.
Fix: restore the route, rule or record, then run a live test in URL Inspection and request indexing for the most important pages.
Verify: the pages return success, carry their content and appear as indexed again over the following recrawls.
Prevent: include a removed URL and a redirected URL among the golden URLs the release checks request, and alert when any live golden URL stops answering with success. If a newly launched site is missing from Google altogether, start with a website not showing on Google.
Addresses invented by scripts, parameters or malformed links
Some entries were never pages: a relative link that doubles a folder, a script that writes a URL from a template variable, an old tracking pattern. The URL parameter audit collects such patterns from crawls, logs and Search Console and stops the site creating them.
Fix: find the link or script that generates the address and correct it; the address itself can go on answering not found, which is the honest response for something that never existed.
Verify: a crawl no longer finds links to the pattern, and new examples of it stop appearing.
Prevent: add new patterns to the parameter register and review script-generated links when templates change. If such addresses answer with an error rather than not found, the page on server errors in Search Console covers that case.
Prevention: removing and moving pages on purpose
Most not-found trouble comes from removals and moves that nobody decided. A short rule prevents it. Every page that ends gets a written decision: keep with a notice, redirect to a genuine equivalent, or remove with a not-found or gone response, as the expired pages guide sets out by page type. On the day of the decision, links and sitemap entries change with it. Every address that changes goes through the URL change procedure and lands in the redirect map. And before anything is removed, someone checks whether other sites link to it.
| Moment | Check | Owner |
|---|---|---|
| A page ends or is deleted | Decision recorded; internal links and sitemap updated the same day | Web publisher |
| An address changes | Redirect written, tested and logged in the redirect map | Web publisher with the SEO lead |
| Before a removal | Links from other sites and past clicks checked | SEO lead |
| Every release | Golden URLs include a live, a removed and a redirected address | Developer |
| Weekly | New not-found examples sorted into expected, leaking, losing, mistaken and noise | SEO lead |
Keep honesty in both directions. A removed page that answers with success becomes a soft not-found error, and a redirect of everything to the home page is treated the same way. The not-found response is the correct answer for a page that is gone; the aim is only that it is never the answer for a page that should be there.
A music school sorts its missing pages (illustrative)
Oxtavenny Music School
Oxtavenny Music School, an invented school teaching piano, guitar and singing, reorganises its course pages and sees not-found errors in the Page Indexing report rise from 30 to 410 example URLs in 5 weeks. Sorting the export by value rather than count: 330 are retired term timetables that should stay gone; 44 are old course URLs still named by a footer link and the sitemap; 9 are instrument pages that a local council's activity directory links to, moved without redirects; 1 is the live booking page, answering not found since a release 12 days earlier; and the rest are doubled-folder addresses from a broken relative link in a newsletter template. The team restores the booking page the same day and requests indexing, fixes the footer and regenerates the sitemap, redirects the 9 linked pages to their successor courses, and corrects the template link. The 330 timetables are logged as expected. A month later the booking page is indexed again and the linked instrument pages receive visits through their redirects.
Common mistakes
- Treating the count as the problem. Removed pages answering not found are correct; sort by value, not volume.
- Redirecting every dead URL to the home page. It is treated as a soft not-found and gives the visitor nothing.
- Removing pages but not the links to them. The site keeps sending people and crawlers to dead ends.
- Moving pages without a redirect map. Every changed address without a redirect becomes a loss.
- Ignoring links from other sites. A linked dead URL is the cheapest recovery on the list.
- Missing the live page among the expected entries. Check the examples against the pages that should exist.
