How a 404 works
When a browser or a crawler asks for an address, the server answers with a status code before it sends anything else. A 404 means the request was understood, the address is valid, and there is simply nothing there. The code is the part machines read; the page you see is a separate thing that your site decides to show alongside it.
That split matters, because the two can disagree. A site can print a friendly “page not found” message while quietly returning a success code, which leaves search engines treating an empty page as real content. It can also return a correct 404 with a blank white screen, which is technically right and useless to the person reading it. Both halves need to be correct.
Why 404s matter
Not-found errors are normal and expected. The web is full of mistyped links, old bookmarks and addresses that were never valid, and search engines have handled them for as long as they have existed. A long list of 404s in Search Console is not a penalty and usually needs no action at all.
What matters is which URLs return one. A page you deliberately deleted should 404 — that is the honest answer. A page that moved during a redesign should not, because visitors and the links pointing at it are being thrown away. The difference between those two cases is the entire job. Everything else is noise, and treating the noise as urgent is how people waste a month on a report that was never a problem.
Common mistakes with 404s
The worst is redirecting every not-found URL to the homepage. It looks tidy, but the visitor loses the thing they came for, and search engines often treat an irrelevant redirect as a not-found response anyway, so you get the same outcome with the confusion added.
The second is leaving internal links pointing at dead pages. An external link you cannot control is somebody else’s mistake; a link in your own menu, footer or article text is yours, and it sends both people and crawlers into a wall. The third is shipping the server’s default error screen with no navigation, no search box and no way back into the site.
How to act on it
Pull the not-found list from Search Console, then crawl your own site separately, because the two find different things. Sort by whether the URL ever existed and whether anything still links to it. URLs with real inbound links or past traffic get a redirect to the closest genuine equivalent — never a token one. URLs that were deleted on purpose stay as they are, or become a 410 gone response if you want the removal to be unambiguous.
Then fix the links you own and make the error page earn its place: a plain explanation, a search box, and links to the sections most visitors want. Handling this properly at launch is part of any sensible technical SEO setup, and it costs far less than repairing it after a migration has already lost the traffic.