What does a good 404 page need?

The status code comes first

A page that says "not found" while returning HTTP 200 is a soft 404. To a browser it looks fine; to everything else it is a real page with real content.

• Search engines index it, filling the index with duplicate "not found" pages.
• Crawl budget is spent on pages that do not exist.
• Link checkers report broken links as working.
• Analytics record them as successful page views.

curl -sI https://yoursite.com/this-does-not-exist | head -1

That must return HTTP/2 404. If it returns 200, that is the first thing to fix — and single-page applications produce this by default, because the server returns the app shell for every path.

What the page should contain

1. Say what happened, plainly. "This page does not exist" beats "Oops!"
2. Say what to do. The visitor wanted something; help them get it.
3. Offer search, if you have it.
4. List the main sections. On a small site, list everything — it is often the fastest route.
5. Guess intelligently. If the path resembles something real, offer it.
6. Keep your navigation. A 404 stripped of the site chrome is a dead end.

Handle the other errors too

404 gets attention and the others are left as the server default — a bare white page with a line of text, which reads as a broken site.

One template reading the status code covers all of them, which is far less work than it sounds.

Prevention

• Redirect with 301 when you move content permanently. A redirect is better than a good 404.
• Use 410 Gone for content deliberately removed — it tells search engines to drop it faster.
• Watch the 404 report in Search Console, and your own logs, for patterns.
• Fix your own broken internal links first. They are entirely within your control and are a large share of most sites' 404s.