How does Rankbrain analyse web design?

22nd January 2016



Fridge Words”  by  Jon Fife is licensed under CC BY 2.0

RankBrain was quietly rolled out in the first half of 2015, but it took Google until November to confirm its existence and to provide any insight into what it is designed to achieve.

RankBrain is primarily targeted at improving Google’s ability to infer the intent of search users even if they enter long, conversational questions rather than making use of short, keyword-centric queries. And in the age of mobile search and voice search, this is essential to improving the relevance of the results which are returned.

Another aspect of RankBrain which sets it apart from previous algorithm updates is its machine learning underpinnings. This architecture enables it to automatically refine itself over time, minimising the involvement of Google’s human engineers in the equation over the longer term.

But how does RankBrain analyse sites at a technical level and is it something which website designers will need to take into account in their work in order to ensure that pages are correctly optimised going forwards?

Words, Words, Words

When it comes to the impact of RankBrain on the position of sites on Google’s SERPs, its primary area of influence relates to the syntactic and semantic choices that are made. And since about 15 per cent of the searches that Google handles every day are unique, having never been previously encountered, a fair degree of guesswork and approximation is required in order to deliver information which is pertinent to the searcher’s intent.

This means that RankBrain has a greater interest in the way in which content on a site is worded, rather than in layout or accessibility. There are hundreds of ranking signals used to calculate the position of each page. While RankBrain may not care whether a site is mobile-friendly or replete with pop-ups, Google already has these other areas covered with existing elements of its algorithm.

Offline Learning

Significantly, RankBrain is neither the first machine learning system to be deployed in the search market, nor does it make real time adjustments to the way in which sites are ranked.

Google has confirmed that the learning process takes place offline and uses samples of search data. So as with other algorithm updates it is likely to roll out changes sporadically, rather than in real time. Site owners should therefore focus on pursuing SEO best practice going forwards, especially when it comes to content creation.