HTTP status code

1xx informational: the request was received and processing continues. `100 Continue` and `101 Switching Protocols` are the two a site may emit; neither is a final answer and neither ends the exchange.

2xx success: the request was understood and fulfilled. `200 OK` is the ordinary answer, `204 No Content` says there is nothing to return, and `206 Partial Content` answers a range request such as a video seek.

3xx redirection: the resource is somewhere else, or the copy the client holds is still good. `301`, `302`, `307` and `308` point elsewhere; `304 Not Modified` says the cached copy may be reused.

4xx client error: the request was wrong in some way. `400` malformed, `401` not authenticated, `403` authenticated but not allowed, `404` no such resource, `410` deliberately removed, `429` too many requests.

5xx server error: the request was fine and the server failed anyway. `500` an unhandled fault, `502` a bad answer from an upstream, `503` temporarily unavailable, `504` an upstream timed out.

A status code is the only part of the answer every automated client agrees on. A browser decides whether to cache, a proxy decides whether to store, a link checker decides whether the link works, and a search engine decides whether to keep the URL indexed — all from the number.

The page body is for people. It carries no machine-readable claim about whether the request succeeded, so a body that says "page not found" under a `200` is read by every machine as a successful page with that text on it.

That mismatch has a name: a soft 404. It is common in applications that render the not-found view in the browser without the server ever being told the route was invalid.

The rule that avoids all of this is short: decide the status on the server, and let the page describe it rather than replace it.

The entry URL's code decides whether the rest of the scan is possible at all. A `5xx` or a connection failure there means nothing downstream can be measured, so it is reported as a state rather than as a list of findings.

Link checking produces the widest spread. Internal links resolving to `404` or `410` are broken; `403` usually means a protected area or a bot filter rather than a dead page; `429` means the check was rate limited and says nothing about the target.

Redirects are followed and counted. A single `301` is normal; a chain of several is a latency and signal-dilution problem, and a chain that returns to its own start is a loop that never resolves.

Codes seen on subresources — images, scripts, stylesheets — matter differently: a `404` on a stylesheet changes what the page looks like, not whether it is indexable.

`curl -I https://example.com/some-path` prints the status line and headers without downloading the body. Add `-L` to follow redirects and `-v` to see each hop's code in order.

In a browser, the Network panel lists the status of every request the page made, which is the fastest way to find a subresource failing under a page that renders correctly.

Check a URL you know is invalid on each distinct part of the site. Applications frequently return the right code for static routes and `200` for application routes, and the difference only shows when you look.

Search Console's Page indexing report groups URLs by the outcome Google observed, which is where a status problem that only appears to crawlers becomes visible.