Client says the 404 page was redesigned three times to match brand colors, yet more empty links appeared in search—is the problem really the visuals?
A good-looking 404 page and whether search engines treat it as an empty link are two different things. What really determines the outcome is the status code the server returns for a nonexistent URL: only a 404 or 410 clearly tells crawling systems that the page is gone; returning 200 with a line saying “Page not found” is what is commonly called a soft 404, and no matter how polished the visuals are, it will still be crawled as a normal page. Based on 2026 project delivery experience, after a redesign, product delisting, or section consolidation, this item should be included in acceptance checks—not investigated only after a batch of empty links appears in search.
Why does a better-looking 404 page sometimes come with more empty links?
When crawling a URL, the system reads the HTTP status code first and then looks at the page content. The trouble with a soft 404 is that the status code says “normal” while the page provides no useful information: crawling systems can easily classify it as a low-quality empty page, and once there are enough of them, search results show a batch of links with similar titles and thin content. So this is usually not a design-file problem, but a misalignment among server responses, template rules, and the cache layer.
The triggers are very specific: old sections are deleted outright during a redesign; a product detail template remains accessible after the product is delisted; pseudo-static rules pass nonexistent parameters to a listing page; a CDN caches the error page as 200; or all incorrect URLs are redirected to the homepage, so the final landing page becomes the homepage. A visitor coming from an old brochure, old email, or external platform sees “Page not found,” while the crawling system receives a different set of signals.
On delivery: check response headers first, then the page's exit points
A common project constraint looks like this: the budget covers only one generic error page, the timeline is squeezed to two or three days before launch, and the only asset is one brand image. The developer conveniently turns it into a static template, the server returns 200 across the board, and visually nothing looks wrong. One or two months after launch, blank links with similar titles begin to appear in search; the client first suspects a hack, but in fact the status codes were not matched. Rework requires changing server rules, CDN caching, and the sitemap at the same time, with an experience range of 1-3 business days, and dead links must be resubmitted—clearly more costly than testing once before launch.
It is best to fix the check order: first look at the response headers, which can rule out soft 404s; then look at the page's exit points, which can rule out “the status code is present but visitors cannot move on”; then check the entry sources—whether links in the sitemap, internal links, and old emails still point to deleted URLs; finally, check the actual test results after clearing the cache. We usually randomly sample 10-20 nonexistent URLs before launch, record the status codes and landing pages in the acceptance checklist, and use them directly for comparison during rework.
The pass line: three checks are enough
The pass line for a 404 mainly depends on a correct status code, a page that can return, and entry points that do not mislead. If any one of these three is missing, the page is only decorative no matter how complete it looks.
- Status code: Sampled nonexistent URLs return 404 or 410, not 200.
- Page exit points: There are entry points back to the homepage, site search, or contact information, and they do not redirect automatically.
- Cache consistency: The CDN and origin server return the same status code for the same URL.
Four treatments you can verify: soft 404, 404, 410, and 301
These four are not aesthetic choices but treatment strategies, and the difference lies mainly in change effort and handoff capacity. Below is a list based on experience ranges in common projects; for any specific site, actual test results still take precedence.
- Soft 404 (returns 200): Low change effort, commonly live within half a day, but easily indexed as an empty page; suitable only for internal test environments and not recommended for public sites.
- True 404: Low configuration cost, commonly half a business day to one day (experience range), suitable for temporary delisting, mistyped URLs, or cases with no replacement page for now; the page should provide exit points.
- 410: More explicit semantics and configuration effort close to 404, suitable for content that is confirmed permanently deleted and will not be restored; not every site must use it.
- 301: Suitable for section consolidation, product model changes, and directory migration; mappings need to be planned one by one, commonly taking half a day to 2 business days (experience range), and incorrect mappings make bad redirects more impactful.
The selection order is also simple: if an old URL has a corresponding new page, prioritize 301; if there is no replacement page, use 404, and if it is confirmed permanently deleted, use 410; for temporary out-of-stock, do not rush to 301 to the homepage, and do not batch-redirect all incorrect URLs across the site to the homepage—that misaligns status codes and landing pages, and visitors cannot find the original content.
When it applies and when it does not: first determine whether this is your scenario
The cases suitable for including 404 in acceptance checks are fairly clear: official website redesigns, section consolidation, product delisting, domain or directory migration, old brochures and old emails still in external use, and dead links already appearing in the sitemap. In these scenarios, a 404 is not just “present and good enough”; it must be handled together with 301s, the sitemap, and internal links.
The cases that do not need a complex solution are also clear: pure single-page showcase sites, small sites with no historical legacy links, and small sites whose content is rarely deleted or changed do not need complex interactions for 404s; internal preview environments should not be accepted against public-site standards either. One boundary that can be quoted on its own: A 404 page is responsible for informing and guiding, while a 301 is responsible for carrying traffic forward; the two cannot replace each other; when there is no replacement page, do not use a 301 to jump directly to the homepage.
Common questions
What happens if a 404 page returns 200?
It will be crawled as a normal page, while the page provides no useful content, easily creating a soft 404; if enough of them accumulate, a batch of similar empty links appears in search, and cleaning them up takes far more time than fixing the status code before launch.
Should I choose 404 or 410?
For temporary delisting or mistyped URLs, 404 is more common; for content confirmed permanently deleted and not to be restored, 410 can be used. Both should retain return entry points, and there is no need to repeatedly change configuration just for the semantic difference.
Should a 404 page redirect to the homepage automatically?
Automatic redirection is not recommended. A forced redirect misaligns the status code and the final landing page, and visitors do not have time to return to where they were; making the return to the homepage a clickable entry point is more reliable than automatic redirection.
If CDN caching is enabled, are the tested status codes still accurate?
Not necessarily. Edge nodes may cache an old response, so even after the origin server is changed, the public network may still return 200. During acceptance, clear the cache and test from a different network, and check the CDN's caching rules for error pages.
If the official website has just been redesigned or a product has been delisted, you can first sample 10-20 nonexistent URLs, check the response headers, and then decide whether to use 404, 410, or 301. This applies to public sites with historical legacy links; internal preview sites do not need to be accepted against the same standards. After the status codes are corrected, indexed empty links, together with sitemap updates and dead-link submissions, commonly decrease gradually over 2-6 weeks (experience range); during this period, there is no need to repeatedly change the page design.
-
Drone Accessories Company Website DevelopmentIncorporating gray as an accent with the pr ...
-
Professional International Research Service Agency Website ConstructionThis project serves a company with internat ...
-
The Construction of Group Websites for Asset Operation and Digital ServicesThis project is to create a website for a c ...
-
Thermal Test Equipment Website ConstructionFounded in 2022, this tech company focuses ...
-
Client uploaded the raw corporate video directly to the website—sales says customers want the full film, and mobile visitors see it spin—should we compress it first or switch hosting?
Date: Nov 17, 2026 Read: 14
-
When people search our company name on Baidu, the top result is a “thank you” page after form submission that only says submission successful—does that mean the official website has a problem?
Date: Nov 16, 2026 Read: 18
-
If a client uploads a certificate photo and says the text is already in the image, whose problem is it when a screen reader cannot read it?
Date: Nov 15, 2026 Read: 24
-
The client deleted a departing colleague's backend account, but their phone number can still reset the password—is that really cleaned up?
Date: Nov 14, 2026 Read: 26
-
A client says competitors’ websites let visitors pan around their workshops. Our workshop is still in production and cannot be cleared out—should we still shoot the panorama?
Date: Nov 13, 2026 Read: 25




