20

Since late 2011, Stack Exchange has stopped using the rel=nofollow attribute on links from posts with a high enough score. However, the precise details of what counts as "high enough" were not disclosed at the time "to discourage gaming of this feature."

While I can sort of sympathize with this view, I think it should be about time to disclose some general information about how the threshold is calculated and what factors affect it, so that we can more meaningfully discuss policy issues related to it.

In any case, even in the absence of an official statement, it's pretty easy to get some idea of how the threshold is set simply by looking at posts with links in them. I like to use a lot of links in my answers (most often to Wikipedia), so I figured I'd start there. Some initial observations include:

  • The score threshold on SO seems ridiculously high. These two answers, with score 43 each, don't have nofollow, but my next highest scoring answer on SO, with score 15, does. You'd think that anything that qualifies for a Nice Answer badge should be good enough not to need nofollow, but apparently that's not the case.

  • On other sites, the threshold seems a lot more reasonable. I have several answers with a score of just 5 on other SE sites that don't have nofollow, although others do have it. Haven't found any un-nofollowed answers with a lower score than that yet, although I can't rule out the possibility that there might be some.

  • It's not just about the answer score — question score (edit: or more likely view count) seems to matter too, even for answers. For example, this answer on SO with a score of 13 (by a lower-rep user, but to a question with score 3650) doesn't have nofollow, even though my score 15 answer above does. That might also be the reason why this answer on math.SE with score 8 has nofollow, even though this answer on gamedev.SE, with the same score, does.

As far as I can tell for this very cursory examination, the main variables, at least for links in answers, seem to be site, question score view count and answer score. I didn't find any answers where only part of the links would have nofollow, suggesting that the target site doesn't matter (with the pretty obvious exception of links to SE sites, which are never nofollowed), and if post age or user rep enter into the calculation somehow, I didn't notice any obvious sign of it.

If anyone has more detailed information on how the decision to use or not use rel=nofollow is calculated, please share it. Also, I didn't (so far) look at links in questions or comments at all, so any information on those would be most appreciated.


Edit: The idea that I might be confusing the effects of question score and view count was suggested in the comments, so I decided to look into it. Here's one data point:

  • Answer A (from above): answer score 15, question score 30, views 521, answerer rep 17.3k, posted in Oct 2011: links do have nofollow
  • Answer B: answer score 6, question score 7, views 178,014, answerer rep 2.6k, posted in Feb 2012: links don't have nofollow

So the commenters seem to be right: view count matters, and based on the observations so far, it's possible that question score does not matter for answers.


Edit 2: Tip: If you're using Firefox and would like an easy way to see which links are nofollowed and which are not, add the following rule to chrome/userContent.css in your Firefox profile directory:

@-moz-document domain(stackexchange.com), domain(stackoverflow.com),
domain(superuser.com), domain(serverfault.com), domain(stackapps.com),
domain(mathoverflow.net), domain(askubuntu.com) {
    a[rel*=nofollow] {
        color: red !important;  /* make nofollow links red */
    }
}

Note that you'll need to restart Firefox for the changes to take effect. Once you do, you may (or may not) be surprised by how many red links you'll see.

Alternatively, you can install this user script (raw link to install), which does the same thing. It has been confirmed to work on Google Chrome and on Firefox using GreaseMonkey.


Edit 3: I just noticed something curious: out of the answers to this question on physics.SE (score 9, 328 views), this one (score 8, poster rep 4.9k) has nofollow, while this one (score 5, poster rep 495) doesn't.

The only obvious explanation I can think of is that the non-nofollowed link is to Wikipedia, while the nofollowed link is to a random not-so-huge website (cirris.com). So it seems that, at least in some cases, the target of the link does matter, after all.

7
  • 1
    @Ilmari Did you consider question view counts in your examination? Nov 29, 2013 at 4:45
  • Yeah, it would seem to be pretty associated with views from the smattering you've posted. The +15 answer you keep citing only has 500 views and change. The other ones are in the thousands (the +13 score answer on the 3k+ scored question has almost 100k views).
    – jmac
    Nov 29, 2013 at 7:38
  • @jmac, Jonathan: That's a good point. I guess I'd need to find some questions with a low score but high views (or vice versa, if there are any) to test it. Or does the system track views separately for answers too, and if so, how do I find out what the view count is? Nov 29, 2013 at 15:12
  • 1
    Views are not tracked separately for answers.
    – Shog9
    Nov 29, 2013 at 16:04
  • 1
    FWIW: there are currently 322182 posts without nofollow links on Stack Overflow. That's just shy of 2% of all visible posts.
    – Shog9
    Nov 30, 2013 at 4:20
  • Is your current edit something that the community needs to address in the answers below?
    – user102937
    Dec 16, 2013 at 18:58
  • @TheGrinch: Just trying to collect what information I can, based on observed behavior. I suppose I could move most of that stuff to another self-answer. Unless, of course, an SE employee did decide to post an authoritative answer... (not that I really expect that to happen, given Shog9's answer below). In any case, sorry for bumping the post. Dec 16, 2013 at 19:02

2 Answers 2

15

I've intentionally avoided looking up the exact criteria for this before answering, because it doesn't matter. If you're just curious, the original discussion covers a number of different criteria that were considered for this, along with the pros and cons of each.

In my opinion, the best argument for ever omitting nofollow from user-submitted links is found in Brian R. Bondy's blog post on the topic:

Abusing nofollow means that the sites that should get credit for good content no longer are getting credit for good content. This will in turn mean that you will receive search results that aren't the best possible ones.

I think we can all agree that when a given site is being referenced often on Stack Overflow, search engines should be taking this into account when ranking search results for related terms. And I hope we can all agree that someone deciding to scatter links to their sites across well-regarded questions on Stack Overflow shouldn't instantly grant them higher ranking in search results.

Sadly, I've had plenty of opportunities to observe unscrupulous organizations and individuals attempting to achieve the latter. Not to get into details, but there are certain product names that are permanently blacklisted on SO right now because of their owners' persistence in this sort of abusive behavior.

I strongly believe we have a duty to do everything in our power to prevent this sort of abuse from ever becoming effective on these sites.

We owe this to the folks who are participating in good faith, to those excellent resources on the 'Net written for the benefit of their readership and not as shameless attempts to garner click-throughs, and to the population of The Internet at large who struggle daily to find good information amid the overwhelming noise. These sites exist to make good information easier to find on the 'Net - we must never lose sight of this goal.

As I said above, the exact method for omitting nofollow from links doesn't matter. What matters is only that it is effective in its goals: allowing clearly good links to be followed while preventing anything in doubt from being trusted.

To this end, if you find a post that clearly can be trusted whose links are nofollowed, report it. Keep a list. If we can adjust our criteria such that more good links are followed, we'll do so.

By the same token, report any sketchy links that don't have nofollow on them! You're trying to reverse-engineer the logic used for this, so you can bet someone else is too... With nefarious intent. If someone does manage to figure out a loophole such that they're able to post or edit in promotional links without the "nofollow", we need to fix that ASAP.

5
  • 3
    Thanks for the response. Just one question: when I do come across posts that IMO shouldn't have nofollow but do, how should I report them and to whom? And does that include my own posts? (I'm pretty sure I've never posted spam on SE...) Nov 29, 2013 at 23:23
  • Also, here's one answer with a nofollowed link that probably shouldn't be. ;) Nov 29, 2013 at 23:27
  • You can start a list in an answer here, if you like, @Ilmari. And that answer has attracted a suspicious number of votes in the past 7 hours for a question that's barely cleared 100 views... Author has an honest face though, so maybe that should count for something ;-)
    – Shog9
    Nov 29, 2013 at 23:46
  • OK, here's a few to start with. I'll try to remember to add more as I come across them; most of these are from going through my own top 50 answers on a couple of sites. Nov 30, 2013 at 1:09
  • Maybe the rep of the person that did the edit to add the link, and also the rep of OTHER people that also added links to the same site should be taken into account. Sep 20, 2014 at 16:13
4

Per Shog9's suggestion above, here's a random list of posts that seem legitimate but still have nofollowed links. I'm mostly just adding posts here as I come across them while browsing (with the user CSS hack described in the question above), so don't expect this list to be in any way exhaustive or logically ordered. I also foresee it getting pretty long pretty quick, but what the heck, let's give this a try...

Examples of non-spammy posts with nofollow:

Some from my own list of highest-scoring answers on SO:

...and from other sites:

Some recent lower-scoring ones by myself. These have a fairly low answer score, but you'd think a user with 17.3k rep could probably be trusted not to be a spammer.

I could keep adding these by the dozen, but that would be boring. Here's the full list, many of them have links and most of those seem to be nofollowed.


Edit: I just realized that even the highest scoring question on MSO (score 1270, 43k views) is nofollowed. In fact, as far as I can tell, it looks a lot like all external links from MSO have the rel=nofollow attribute. Surely at least that should be fixed.

More examples:

...and one more:

1
  • Ok, so I checked - meta sites are explicitly excluded from the logic that removes nofollow. No user-submitted links on meta sites should ever not have nofollow.
    – Shog9
    Nov 30, 2013 at 4:10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .