October 28th, 2015

Google have announced (seemingly via Bloomberg) an addition to their Hummingbird search algorithm, called RankBrain. They say that it is a form of AI or machine learning that has been tested in their search algorithm for the past few months to help with answering never before seen queries, with what appears to be very pleasing results. There isn’t a huge amount of information out there on RankBrain and Google are playing their cards very close to their chest, so much so that they have not even commented on the update on their Webmaster Central blog. So here is what we know:
RankBrain works in conjunction with Google’s major search algorithm, Hummingbird. Like all the other updates that you may have heard of including #Mobeilgeddon and the famous Panda and Penguin, RankBrain is a bolt on to the core search algorithm. However, unlike other updates which require considerable amounts of human interaction and rule setting, RankBrain’s AI creates a much more fluid ecosystem. According to Blomberg it ‘uses artifical intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand.

So what does that actually mean?

Google’s search algorithm has been created by humans according to trends or patterns that those engineers have identified via the data available to them. It is a very complex task and one for the in demand data nerds. The issue is that this leaves a very rigid structure that can only then be truly updated by a human going in and altering the algorithm. It would appear that RankBrain is the official manifestation of what many SEOs believe has been written on the walls for a while now: AI or machine learning that is constantly updating the way in which Google ranks websites according to search results.

A major factor behind the release of RankBrain is the way that we search. Google claim that around 15% of our 3 billion searches every day are ones that Google has never seen before. This percentage is lower than the 20/25% of previous years but still equates to nearly half a billion searches a day that Google needs to return the most appropriate result for. RankBrain is supposed to be able to more accurately identify the core meaning, including synonyms and intent, of these new searches and provide better results. In fact the Bloomberg articles states that compared with human judgement (Google engineers), RankBrain was 10% better at returning results to these new queries. It certainly feels as though RankBrain is an addition to the original Hummingbird update and may be a continuation of Google’s ability to return results for the more colloquial voice activated search.

An ever evolving algorithm

Google have said that after their period of testing RankBrain has established itself as the third most important ranking factor, although they neglect to mention which factors take the gold and silver medals. Remember that Google takes into account hundreds of ranking signals, of which RankBrain is now one of them.

Danny Sullivan at Search Engine Land ran a couple of tests for RankBrain with mixed results, although he also proclaimed his faith in RankBrain’s ability to improve search results. With an algorithm that is constantly learning we can expect it to become better and better at its role in the search engine algorithm as a whole. As a result, improving Google’s search experience automatically, rather than having to wait for the Google engineers to do their stuff.

Will RankBrain change SEO?

It shouldn’t. The vast majority of Google updates should not change the strategies of White Hat SEO, it is usually only those on the other side of the fence that complain about Google moving the goal posts. Content will continue to be a major focus for search marketers and we do not envisage RankBrain changing this. If anything, RankBrain will place even more focus on a website’s ability to provide high value content and therefore be returned for longer tail search terms.