As a well connected SEO I digest a lot of publications from the web and I try to limit my opinion to factual results either from real world feedback or by controlled tests. Google is constantly evolving and improving itself to render the best search results possible, or at least better search results than the competition.
Considering where Google was with regards to just hardware in 1999, things certainly keep changing:
On Monday SEO Moz published a small test they did to gauge the importance of keywords in the anchor text of links. The test is discussed in detail over on SEO Moz but the result was rather straight forward.
In a nutshell they took 3 new sites, randomly equivalent, and tried to build some controlled links to the sites using three different approaches:
- Build links with just ‘click here’ text
- Build links with the same main keyword phrase
- Build links with random components of the main keyword phrase
Obviously the test is a bit broken, because if you don’t have existing keyword relevance for a phrase, you should build relevance with keywords in the anchors. When Google is sorting out who will be ranked #1 for a site dealing with candies, the site linked to with relevant keywords should always rank higher than a site with links like “click here” or “this site” which aren’t relevant. The only exception would be in a situation where the links seem excessive or ‘spammy’ and may result in Google not considering any of the similar links for relevance.
Outside of a clean test environment we know the best results would be a blend of all three types, with a bit of brand linking mixed in to avoid losing focus on brand keywords. A well established site with a healthy user base will constantly be establishing brand due to all the time on site and click-through traffic for that brand.
ie. If I search for “Sears” and click on the first link only to find it’s a competitor, I’d hit back and find the right link to click. In most cases Google’s watching/learning from the process, so brand links aren’t going to be a necessity after a site is quite popular, and the % of brand links wouldn’t need to be much at all.
Kudos to SEOMoz for publishing some of their SEO test info regardless of how experimental it was. We’re constantly putting Google’s updates to the test and it’s often very hard to publish the results in such a clinical fashion for all to see. We will always make an attempt to blog on the topics we’re testing but it’s still on the to-do list to publish more of the data.