An HTTP response indicating that the requested resource could not be found.
A 404 response means the server could not find the resource requested at that address. It does not, by itself, say whether the resource once existed or whether the absence is permanent. A mistyped URL, a broken internal link and an intentionally removed page can all produce a 404.
The response code and the page shown to the visitor are separate things. A business can return a genuine 404 while displaying a helpful explanation, a search box or links to useful sections. A visually attractive error page should still communicate the correct HTTP status to software requesting the URL.
Not every 404 needs a redirect. If a useful page moved, its old address should usually lead to the relevant replacement. If the URL is meaningless or the content is genuinely gone with no suitable successor, an honest not-found response may be appropriate. The goal is to repair broken customer journeys, not force every report to show zero errors.
Takeaway: Prioritize broken links that affect real visitors and important pages. Keep a useful not-found page, but do not send all missing URLs to an unrelated homepage just to hide the status.
Example
A menu link points to /contact-us/ after the contact page moved to /contact/. Fixing the menu and mapping the old address to the actual contact page resolves a real problem. A random misspelling typed by a visitor does not necessarily require a permanent redirect rule of its own.
What to check
For a reported 404, check the exact URL, where it is linked and whether an intended replacement exists. Test the HTTP response as well as the visible page. A message saying page not found with a 200 response can create a soft 404 problem.
For more detail, see MDN 404 reference.