"How does Google consider site-wide backlinks?"

Matt Cutts from Google discusses how Google handles sitewide links from both an algorithmic and a manual standpoint.

"Are sitewide links effective for SEO?"

"Are site-wide backlinks considered good or bad by Google? Or do they just count as 1 link from the whole domain?"

That's an excellent question.

Let's tackle it from both the algorithmic standpoint and from the manual standpoint.

On the algorithmic standpoint. Typically I've said before that if we have keywords, the first the keyword counts some, the next keyword counts a little bit but not as much, the 3rd keyword not as much. So even if you do keyword stuffing and even if you throw a ton of keywords, at some point it becomes an asymptotically diminishing return and it doesn't really help you anymore.

You can imagine the same sort of things, you know, if we see a link from a domain, we might count it once but if we see 50 links from a domain we still might choose to count only once.

So on an algorithmic side we do a pretty good job of compressing those links together.

But then there's also a manual side. So imagine that you have a Polish website and then you see a site-wide link in English talking about "rent cheap apartments". You know, to a regular person that looks pretty bad.

Certainly it does happen that you have sitewide links—maybe have a blogroll or something like that—but if I were a manual web spam analyst who was sort of doing an investigation and got a spam report that in your English site you've got a sitewide Polish link that looks commercial, off-topic, of low quality, or spammy, then that can affect the assessment on whether we want to trust the outgoing links from that site.

Site-wide links do happen—it's completely natural, you might have a
privacy policy, copyright policy, all that sort of stuff. But when we see sort of irrelevant, off-topic links, or even spammy links, site-wide, then that's the sort of thing that does make a spamfighter's ears sort of perk up a little bit.

I think we handle it very well on the algorithmic side but, you know, we do manual investigations as well.

Hope that helps.