How to Improve 404 Errors: Find, Prioritize and Fix Broken URLs

What a 404 error means

404 Not Found is an HTTP client-error response indicating that the origin server cannot find a current representation of the requested resource, or is unwilling to disclose that one exists. The status does not establish whether the absence is temporary or permanent. RFC 9110 reserves 410 Gone for cases where permanent removal is known.

A 404 is not automatically an SEO defect. Deleted products, expired events, malformed URLs and links typed incorrectly can all produce legitimate 404 responses. Improvement means reducing avoidable dead ends while preserving accurate status codes for resources that really are absent.

The harmful cases are usually active broken internal links, valuable URLs removed without a migration plan, soft 404s and high-demand pages sent to irrelevant destinations. Those problems affect navigation, link equity, conversion paths, indexation signals and crawler efficiency.

Prioritize 404 errors by value and cause

Do not judge a site by its raw 404 count. A large site can accumulate thousands of harmless requests from typos, obsolete external links and crawler-generated URL combinations. Rank each URL by recoverable value and whether the site itself is causing the request.

A practical scoring model can assign points for internal inlinks, organic landings, conversions, referring domains, sitemap inclusion and repeated Googlebot requests. Work from the highest score downward rather than clearing reports alphabetically.

Find the source of each broken URL

Combine several datasets because no single report reveals every cause. Crawl the site to find internal links that resolve to 404. Export the source page, broken destination, anchor text, link location and redirect path. Check Google Search Console for not-found examples, but remember that discovery does not prove that a URL remains internally linked.

1. Crawl internal links: Include navigation, pagination, canonicals, hreflang references, image links and XML sitemaps.
2. Review analytics: Find 404 URLs receiving landing sessions, conversions or repeated user visits.
3. Check backlinks: Identify broken destinations with authoritative or relevant referring pages.
4. Analyze server logs: Segment requests by users, Googlebot, Bingbot and other verified crawlers. Look for recurring patterns rather than isolated hits.
5. Inspect templates and databases: A single defective product, location or CMS template can generate thousands of broken URLs.
6. Test the response: Confirm the initial status, every redirect hop and the final rendered result.

For recurring patterns, fix the generator rather than creating one redirect at a time. Examples include a missing trailing-slash rule, incorrect relative links, deleted locale folders and case-sensitive paths.

Use this 404 decision framework

Choose the response based on user intent and resource equivalence, not on a desire to eliminate an error report.

1. Does the requested page work at another permanent URL? Use a direct 301 redirect to that equivalent URL and update internal links.
2. Is the move genuinely temporary? Use a 302 redirect and retain the original URL in site references where appropriate.
3. Was the page removed, but a close substitute satisfies the same intent? Consider a 301 only if the substitute is materially equivalent. A product category is not automatically equivalent to every discontinued product.
4. Is permanent removal known and no substitute exists? Return 410. This communicates removal more explicitly than 404.
5. Is the URL unknown, mistyped or unsupported? Return a real 404.

Do not mass redirect deleted pages to the homepage. Google explicitly warns that irrelevant redirects can confuse users and may be treated as soft 404s. Redirect chains should also be collapsed so the old URL points directly to the final destination.

Implement fixes without creating new technical debt

Repair links at their source

Changing an internal link is better than relying indefinitely on a redirect. Update navigation, body links, canonicals, hreflang annotations, structured-data URLs and sitemap entries. Search CMS fields and reusable templates for the old path.

Map migrations by intent

For a site migration, build an old-to-new URL map before launch. Match products, services, articles and local landing pages individually where possible. Preserve query intent, language and geographic relevance. Test representative URLs and high-value pages in staging, then verify production responses after release.

Handle inventory and local pages carefully

A temporarily unavailable product may remain live with accurate availability information if it is expected to return. A permanently discontinued product can redirect to a true successor, but not an unrelated bestseller. Closed location pages may redirect to a replacement location only when it reasonably serves the same audience. Otherwise, explain the closure on a live page for a useful transition period, then retire it deliberately.

Keep redirect rules documented with an owner, reason and review date. Broad regular-expression rules should be tested against a sample of valid and invalid URLs before deployment.

Detect soft 404s and JavaScript failures

A soft 404 occurs when a page appears to say that nothing was found but returns 200, or when an irrelevant redirect behaves like an error page. Search systems must infer that the content is missing. Google’s crawl-budget guidance says soft 404 pages can continue consuming crawl resources, while a true 404 strongly signals that the URL should not be recrawled.

Test both the raw HTTP response and rendered output. This is essential for single-page applications. A server may return the same 200 shell for every route while client-side JavaScript displays a not-found message. Configure server-side routing, dynamic rendering infrastructure or framework-level response handling so invalid routes emit meaningful status codes.

• Verify status codes without relying only on what a browser displays.
• Search rendered text for empty-state or not-found templates.
• Confirm that an error route is not canonicalized to an unrelated live page.
• Test with JavaScript enabled and disabled where the architecture permits.
• Check whether blocked scripts or failed API calls incorrectly turn valid pages into apparent errors.

Build a useful custom 404 page

A custom 404 page should return 404 while helping a person continue. Use a clear explanation, normal site navigation, a working search function and a small set of contextually useful destinations. Preserve the site’s visual identity, but keep the page lightweight enough to work when other services fail.

For an ecommerce site, offer category search rather than automatically redirecting every missing product. A local business can surface its location finder, contact route and primary services. A publisher can offer topic hubs or recent resources. Avoid overwhelming users with an entire sitemap or implying that the missing content still exists.

Measure the page as a recovery interface. Useful signals include search usage, onward-click rate, exits and successful conversions after a 404 encounter. Do not add index-oriented structured data that describes content unavailable on the visible page. The error template itself should not be submitted in a sitemap or made indexable through inconsistent canonical signals.

Monitor 404 health with actionable KPIs

Separate outcome metrics from raw counts. Total 404 requests can rise because bots probe random paths, even while site quality improves.

• Broken internal URLs: Count unique 404 destinations linked from crawlable pages. The target should be zero for important templates and navigation.
• Internal 404 inlinks: Track the number of source links, not only destinations. One missing asset can have thousands of inlinks.
• Valuable 404 landings: Monitor sessions, conversions and referring domains reaching unavailable URLs.
• Soft 404 rate: Test pages returning 200 that have thin, empty or not-found output.
• Redirect quality: Measure direct one-hop mappings, chains, loops and destinations that fail intent review.
• Sitemap validity: Require submitted URLs to return 200 and be intended for indexing.
• Crawler demand: Use logs to track repeated requests to high-value missing URLs and malformed URL families.

Set alerts for sudden increases after deployments, CMS releases, taxonomy changes and migrations. Review high-value 404s weekly on large commercial sites, while stable smaller sites may use a monthly cycle.

Turn selected 404s into SEO recovery opportunities

Broken URLs with strong backlinks can reveal content demand. Restore the original resource when it remains useful, consolidate it into a stronger topical hub or redirect it to a genuinely equivalent page. Contact important linking sites when correcting the source URL would be better than maintaining a redirect.

Use this process for content decay remediation. Compare the retired page’s historic query theme, links and conversions with the current content graph. A recovered statistics page, technical reference or original dataset can regain usefulness and attract natural citations. Link it from the appropriate hub and related spokes rather than leaving it isolated.

For AI Overviews, AI Mode and other answer systems, a missing or nonindexed page cannot reliably function as a supporting source. Google states that pages shown as supporting links in its AI features must satisfy foundational Search requirements, including indexability and snippet eligibility. Clean status handling therefore supports retrieval, but it does not guarantee selection or citation by any answer engine.

A higher-risk tactic is redirecting expired domains or unrelated historical pages solely to transfer authority. Relevance, ownership history and user expectations can be weak, and manipulative implementations may create spam or trust risks. Prefer recoveries tied to a real resource, audience and editorial purpose.

What is proven, accepted and still uncertain

Established by standards or official guidance

404 means the requested representation was not found, while 410 communicates known permanent removal. Google recommends 301 for permanent moves, 302 for temporary moves and real 404 or 410 responses when no replacement exists. Google also advises against irrelevant homepage redirects and requires meaningful status handling for JavaScript sites.

Strong practitioner consensus

Technical SEO teams generally prioritize internally linked 404s, sitemap errors, URLs with traffic or backlinks and failures created by templates. Ahrefs and Screaming Frog guidance aligns with repairing source links and redirecting only when a relevant replacement exists. Current Reddit discussions echo these priorities, but community reports are anecdotal rather than causal proof.

Still contextual or uncertain

There is no universal safe number of 404s and no fixed time in which every missing URL disappears from search systems. Crawl frequency depends on site demand, history and URL discovery. Research shows that broken links and soft 404 behavior are widespread, but prevalence figures from a dataset should not be treated as performance benchmarks for every site. The business value of restoring a page also depends on intent, link quality, conversion potential and maintenance cost.