In short
A 404 page should say plainly that the page was not found, offer the most likely routes onward — search, the main sections, popular destinations — and return an actual 404 status code. The clever illustration is optional; the routes out are not. Most visitors reaching one arrived from a stale link and still want what they came for.
What the visitor is doing
They clicked something, or typed something, expecting a specific thing. They now have nothing.
The page's job is to get them to what they wanted, or to the nearest useful alternative. Everything else — the illustration, the joke, the brand personality — is secondary and occasionally counterproductive, because a page that is mostly a joke communicates that you find the failure amusing.
A little personality is fine. It should not occupy the space where the help goes.
What it needs
A plain statement. "This page doesn't exist" or "We couldn't find that page." Not an error code as a headline; most people do not know what 404 means.
Search, if you have it, prominent and focused. This is the single most useful element on the page.
The main sections, so someone can navigate onward rather than leaving.
A few likely destinations — your most visited pages, or the ones most people who hit a 404 end up wanting.
A way to report it, small but present. Visitors will tell you about broken links if you make it possible, and that is free QA.
Full site navigation, unlike a landing page. This is exactly the moment someone needs the menu.
Use the URL they asked for
The URL contains information, and most 404 pages ignore it.
If someone requested /insights/how-to-design-a-pricng-page/, you know what they wanted. You can suggest the closest match, run the slug through your search automatically, or at least show results from the section they were trying to reach.
Even simple handling helps: if the path begins with a known section, show that section's index and recent items rather than a generic page. A visitor who wanted an article and lands on a list of articles has mostly been rescued.
Get the status code right
The most common technical error is a 404 page returning 200 OK.
That is a soft 404. Search engines are told the page exists and has content, so they may index a set of identical error pages. It also breaks monitoring, since nothing registers as an error.
A missing page must return 404. A page that has been permanently removed and will not return can use 410, which is a stronger signal and can be crawled out of the index faster.
And do not redirect all 404s to the homepage. It looks tidier and it is worse: the visitor loses the information that the page is gone, and search engines treat it as a soft 404 anyway. Redirect only where there is a genuine replacement — that is what a redirect map is for.
Find out why they are happening
A 404 page is a symptom. The useful work is upstream.
Log them, with the referrer. A 404 arriving from an internal link is your bug and should be fixed immediately. One arriving from an external site may deserve a redirect, since someone linked to you and that link is worth preserving.
Check Search Console for 404s Google has found, which surfaces pages that once existed and now do not.
Watch for patterns. A single URL generating thousands is usually a broken reference in a template, an old app version, or a mistyped link in an email campaign — all fixable at source.
Most 404 traffic comes from a small number of URLs, so fixing the top handful usually resolves most of it.
The details that get missed
Make sure it works. A 404 page that depends on a broken layout, or on JavaScript that fails, is a blank page at the worst moment.
Keep it light. It should load fast even when something is wrong.
Do not index it. The page itself should not appear in search results.
Handle other errors too. A 500 page deserves the same care and is more often forgotten, since it appears only when something has gone wrong — which is exactly when you cannot fix it live.
Test it on mobile, where most visitors will encounter it.
The simple version
A clear statement, a search box, your navigation, a few likely destinations, and a correct status code. That covers nearly every case.
The elaborate illustrated 404 is a nice touch if the rest is right. It is not a substitute for the rest being right.
If you are migrating a site and want the 404s handled properly rather than discovered later, book a call.
Common questions
What should a 404 page include?
A plain statement that the page was not found, a search box, your full site navigation, a few likely destinations, and a small way to report the broken link. The routes onward matter far more than the illustration — someone reaching a 404 was trying to do something specific.
What status code should a 404 page return?
404. A common error is returning 200 OK, which is a soft 404: search engines are told the page exists with content and may index a set of identical error pages, and your monitoring never registers an error. Use 410 for pages permanently removed that will not return.
Should I redirect 404s to the homepage?
No. It looks tidier but is worse — the visitor loses the information that the page is gone, and search engines treat it as a soft 404 regardless. Redirect only where a genuine replacement page exists, and let everything else return a real 404.
How do I find out why 404s are happening?
Log them with the referrer. A 404 from an internal link is your own bug and should be fixed at source; one from an external site may deserve a redirect, since someone linked to you. Check Search Console too, and look for patterns — most 404 traffic comes from a small number of URLs.
Can a 404 page use the URL that was requested?
Yes, and most do not. The requested path tells you what the visitor wanted, so you can suggest the closest match, run the slug through search automatically, or show the index of the section they were trying to reach — which rescues most visits.
