Ranking and Authority Issues for Twitter Search

by Aaron Bradley on June 23, 2009

in Social Media

Ben Parr's Mashable Post that Google may launch a microblogging (Twitter) search service has ignited a storm of speculation in the SEO and SMO communities. Perhaps most interesting object of speculation is how Google will calculate authority in order to rank tweets in search results.

Jordan Kasteler has written a thoughtful post on this issue in which he introduces the notion of "InfluenceRank" for social media ranking (at least I haven't seen the term coined before), following a couple of patent applications relevant to social media chronicled by Bill Slawski.

As Parr points out, Google is probably seeking to extend the hub/authority model (Ã la Hilltop) to social networks. There, are, however, unique challenges posed by Twitter. Twitter is somewhat of an untamed frontier from an optimization perspective these days, with very few spam controls and filtering options – almost analogous to the ease of gaming the <meta> keywords tag back in the good old Alta Vista days.

What are the possible signals Google could use to ascribe authority to tweets or tweeters, and how might these open up tactical SEO exploits?

Followers

A very problematic metric, given that inflating a user's Twitter follower count has become a cottage industry in its own right. WeFollow ranks tweeters by topic that way, and it doesn't take too much digging to determine that the stack degrades quickly (check out the mix of legitimate SEO "gurus" and guys and girls you've never heard of in the SEO category).

On the other hand, a tweet from a tweeter who has few followers really seems to lack authority on the face of it. But even at that, there are huge discrepancies in relative follower numbers even between similarly "important" or even active tweeters.

Possible exploit(s)? As suggested, even more aggressive efforts to artificially inflate follower numbers.

Twitter Profile PageRank

Insofar as there's validity to link-based votes for a web page, there's some sense to passing PR value to a profile owner. However, the social media space is different from the realm of web pages, and it may not say much to the quality of the target tweeter's actual Twitter content that there's links to that profile. Furthermore, unlike a web page, there's not much to a profile (excepting below) by which Google can correlate links in the way that they can with target web pages.

Possible exploit(s)? Dedicated Twitter profile link-building efforts. These might prove more successful than web link building efforts, since there's nothing suspicious in building links to a single Twitter profile, where as a surplus of links to only one or two pages on a website (shallow linking) raises alarms.

Frequency (Freshness) and Uniqueness of Tweets

Frequency seems to be a fairly useful metric for a microblogging search engine as one of the primary reasons for having a Twitter search engine is to return highly topical results. However, this needs to be correlated against uniqueness to toss out spammy reiterative messages, mundane retweets and other less-than-useful content.

Possible exploit(s)? Populating tweets with the Twitter equivalent of scraped content. Plus it will be much harder to toss out grammatically or semantically suspicious content, unlike web pages. And, while not really an exploit, those who tweet more will have an advantage over those who tweet less, regardless of quality: signal to noise ratio is difficult to determine in microblogging (from an algorithmic perspective).

Link Click-Throughs

In some ways, links are really at the core of an individual tweet's value. However elegant a writer may be, there's not a lot of depth you're going to be able to push out in less than 140 characters. So the number of times users click on a tweeted or retweeted (see below) link really is a value vote for that microblog entry.

Interestingly, Twitter links are not only nofollow'd, but almost always go through a 302 redirect as a result of a URL shortener – and the same target link can be tweeted as any number of shortened links. In order to truly calculate clicks to a single URL, a search engine would need to follow the redirect path in order to "canonicalize" links themselves.

Two slight digressions here. First, it will be interesting to see how a Google microblog search engine would list links. From reasons ranging to a perceived match in topicality to fear of malware users do take a target URL into account before clicking on it, and it would be far more useful to see the fully-resolved URL rather than the shortened URL in a search query result.

Second, there's an opportunity to bring Twitter into the web page ranking algorithm by looking at Twitter click patterns, the nofollow attribute and 302 issues nothwithstanding. I'd say that 1 million Twitter clicks would be indicative of the value of a target page – perhaps more so than click metrics in the search results themselves. Google is not obligated to respect nofollow, nor to ignore 302 targets. It will be interesting to see if there's some sort of link juice capable of being carried to the Web from Twitter, or whether Google will keep those worlds entirely separate.

Possible exploit(s)? Network clicks, bot clicks, articially "viral" retweets and even bait-and-switch URLs.

Retweets

This seems to me a very potentially useful ranking factor. In order for any Twittersearch results stack to be useful, Google wil have to determine the "canonical" tweet (which, as not all retweets are "RT"-prefixed and may be modified, will take some overhead, as will sorting out the "canonical" target URLs of assorted, related tweets).

Possible exploit(s)? Mass retweeting within advocacy or other self-interested networks.

While it's hard to know how Twitter search rankings will ultimately be determined, I confidently predict a whole new SEO tactical gold rush when a Google microblogging engine appears. Twitter verticals, anyone?

{ 0 comments… add one now }

Leave a Comment

Previous post:

Next post: