Rel=nofollow cheat sheet: when and how you should (not) use it

After Matt Cutts’ long-awaited post on PageRank sculpting, there’s obviously plenty of (controversial) discussion going on about the subject right now, both in the comments to Matt’s post, and on nearly all major SEO blogs and discussion boards. While a good 90% of it is pure crap, there are a few articles and commentaries that do make for an interesting read. If you’re in a hurry and still behind with your reading schedule, make sure you don’t miss these:

• a good summary by Andy Beal;
• Danny Sullivan’s comments on the news and the way it was disclosed to the SEO community;
• Rand Fishkin’s disappointment about, and analysis of, the change;
• Aaron Wall’s rant;
• the inevitable WebmasterWorld thread (well, not exactly a must-read; just some stuff the old-skool SEOs and BBS fans out there will appreciate).

Me, I decided to skip past all the speculation and borderline-paranoid theories that have always arisen (“always” meaning “in the last 200 years or so”) whenever Google decided to change things around (SEO is a universe in perpetual change: deal with it), and give some practical advice on when and how, in my opinion, you should or should not use the “rel=nofollow” attribute. Nothing really new here; just my own little cheat sheet for the do’s and don’ts of nofollow, covering its most common uses and misuses.

Do not use nofollow

On internal links:

• For sculpting your PageRank (i.e., to “save” PR for your other links): ’nuff said.
• For siloing your theme: “theme siloing” is so 2005 a technique (and a couple of things have changed on the Web in the last four years, you know), although I agree it can still prove effective in a few cases; if that’s your case, however, you’ll want to go for a well-designed site structure in the first place.
• To link to pages whose contents you don’t want to show up in the SERPs: there’s robots.txt (or the robots <meta> tag) for that, sonny.
• When linking to non-canonical URLs that are only slightly different from the canonical form (e.g., HTTPS vs. HTTP, additional querystring parameters, and so on): rel=canonical + an XML sitemap + Google’s über-intelligent canonicalization algos will do the trick.
• For “crawl prioritization”: I’m sure lots of high-profile folks will disagree on this one, but I believe there are just too many better/faster/stronger ways to do it than using nofollow. To name but a few: XML sitemaps, HTML sitemaps, RSS feeds, server headers, and maybe the single two most important factors: site architecture (a.k.a. “classic” PR sculpting ) and content freshness.

On external links:

• When linking to relevant, on-topic content from authoritative sources: that’s the single case when linking out is only going to do you good (notice the links in this post?), so please, stop being afraid of “losing” your ridiculous PageRank and just fishin’ Do It. You’ll thank me later.
• To make a reciprocal link appear uni-directional, or to “hide” a network from Google: come on, they’re supposed to be a bit smarter than that in Mountain View.
• When linking to Google: ’cause they won’t dig that (just kidding).

Do use nofollow

On internal links:

• On links to login/registration/shopping cart/terms&conditions/privacy policy pages: yes, it’s an Officially Accepted Google-Approved Best Practice™ –even if, to further mess things up, Engineer Matt added that he would try not to nofollow the links to TOS and privacy pages, just to get some filler traffic from long-tail queries. Go figure!
• When linking to non-canonical URLs that are radically different from their canonical form and when rel=canonical is not an option (e.g., different domain name): the best choice would obviously be not to link at all or do a 301, but alas, we all know how nasty some CMSs can be, and “You Can’t Always Get What You Want” is our first motto.
• When linking to pages with substantially duplicate content, or no substantial content at all: (see above).

On external links:

• On links in user-generated contents (e.g., blog comments), if you just don’t have the time to pass every single link under the microscope, and whenever preemptive censorship moderation is practically impossible: anti-spam plugins like Akismet only catch the more blatant forms of comment spam, and you never know the sneaky shit some people will try to link to from your blog, so go for the nofollow!
• When you don’t completely trust or appreciate the website you’re linking to: whenever you can’t or don’t want to “vouch for” (and lend link juice to) the external page you’re linking to, maybe because you don’t like its “neighbourhood”, or you don’t want your site to be associated with it.
• When linking to spammy pages: I know, rule #2 of white-hat SEO is that you shall never link to spam, ever (guess what rule #1 is?). But you might just happen to want or need to show off on your blog, for all the white-hat SEO world to see, some very negative example: that’s where nofollow comes in really handy.
• On paid links (if you do sell links from your site, that is): better play it safe than risk having your site’s “voting power” nuked forever. Don’t say I didn’t warn you.
• On affiliate links: because nobody wants to see that shit in the SERPs, except the affiliate link spammers that we all despise (especially Google).

On any link:

• Don’t abstain from using nofollow just because you fear that, by doing so, you will end up being profiled as an SEO by Google: too late, pal. If you’re reading this blog, chances are that Google has already “profiled” you (D’OH!), so stop whining and grab your badge.