SEO? 404 pages noindex,nofollow or?

[onlinesource]

Curious on a 404 page for deadlinks, would it be wiser to make that page noindex,nofollow or noindex,follow?

I don't know I don't want to index that page because every time somebody creates a link somewhere that's dead, that's a link and before you know it, you could have 500+ 404 urls. :(

Yet, let's say you once had a link called /apple-pies.html but you decided to not sell apple pies anymore, so mysite.tld/apple-pies.html goes to a generic 404 page that says something like, "Sorry, looks like we lost that page... here are some links to great pages you might love", and even though those internal links may have NOTHING to do with apple pie, wouldn't I want Google to follow those internal links? So wouldn't noindex/follow be best or because the content may or may not relate enough, it's always wiser to go with noindex/nofollow?

[keyplyr]

I use "noindex." Never mind the "follow."

SE's aren't going to index it if they got there with a 404; and not indexed directly unless you have a lot of content on it, which you shouldn't anyway.

I also include a search utility so the visitor can search for the correct page.

[lucy24]

I do belt-and-suspenders. Error pages are located in a roboted-out directory, so a law-abiding search engine can't get there by requesting "404.html" or similar*. But each individual error page also has a "noindex" tag. (This is probably the only situation where a "robots" header is meaningful on a roboted-out page.)

wouldn't I want Google to follow those internal links?

Well, yeah, but surely there are lots of other ways for humans to find those pages? It's a UX disaster if the only way people ever find an attractive page is if they blunder across a link on a 404 page. And if humans can find them, search engines can find them.

Reminder: It's never necessary to say "follow". That's simply the default.


* Due to an error I've never been able to pinpoint, the bingbot periodically requests /directory/gone.html, referring to a specific directory with a custom 410 page. Who knows how often they'd be requesting ordinary error pages if they were crawlable?