Rankbrain & Machine Learning – How they will affect SEO

by Luke Holman on Friday 23 September 2016

There have been many interesting developments this year within search which speak to the new vision and the long term goals of Sundar Pichai’s Google. Referencing the statements, he made during the Google I/O conference, in which he reiterated his intention to make Google an ‘AI first’ company and ‘to put machine learning front and centre of everything they do’. This is a clear diversion away from their mobile first approach of recent years and many of the new initiatives such as bots & voice search will have lasting impacts not only on SEO but the wider digital industry for years to come. Of course, a ‘post’ website world is some way off, there is, however, a piece of technology which is quietly rewriting the rule book of search and the implications of which are impressive, to say the least.


What is it?

This initiative is called Rankbrain, confirmed to exist in October 2015 it is Google’s Artificially intelligent (AI) ranking algorithm. Running on 15% of searches at the time, it quickly became the third most important ranking signal. Recently it was confirmed to be running on all queries, around 2 trillion a year, making it one of the most important pieces of technology built by Google to date. This is a clear example of the sheer confidence that the company has in AI technology, expanding from 15% to 100% of searches in less than 8 months. As Rankbrain develops, on-site signals will also move away from ranking inputs like keywords & snippet optimisation and move towards searcher outputs, such as CTR, short vs long clicks and other user-related metrics. There is seemingly a clear movement towards fulfilling user search intent and moving away from what they deem important, perfectly encapsulating the company’s original philosophy of ‘Focus on the user and all else will follow’.


What does it mean?

For SEO’s this presents a unique set of challenges in both the short and long term. We as an industry need to be ready for what could be a complete overhaul of how sites are ranked in the SERP with little to no warning from Google, putting those who are not prepared for this potentiality in an extremely detrimental position.
Short term, things are still fairly simple, focusing on content will still be a winning formula ensuring that best practice is followed with a larger focus on engagement metrics overall. Optimising for CTR’s with well-written title tags & meta descriptions could mitigate any potential fluctuations that may also arise over time. This is because Rankbrain currently helps with interpreting queries for user intent but as the system is continually learning things could change very rapidly.
Long term, everything else is a lot less clear and unpredictable, even one of Google’s core ranking engineers Paul Haahr has admitted that they 'don’t fully understand Rankbrain'. As more and more of search is handed over to the AI, which can constantly update and adapt itself to new information the less Google will know which specific ranking signals are prioritised in the SERP for a given query. The way this will manifest is completely unknown, with many SEO’s predicting various scenarios such as different algorithms depending on the industry vertical or a different priority of ranking signals dependent on the query, with a seemingly infinite set of combinations.


What can be done?

As the levels of uncertainty increase around specific signals, along with a diminishing amount of information provided by Google, it will be the responsibility of the industry as a whole to seek out those answers. SEO’s will need to hypothesise and test specific models more than ever to ensure that accurate information is continually relayed to clients and fed back into strategy. SERP and competitive comparison will be an ever-essential part of that equation to ensure sites are performing to their maximum potential. It will also become paramount that analysing at scale will be the only way to generate statistically significant results to these questions, and the adoption of machine learning by SEO’s should be welcomed with open arms.
This new reality of search is fast approaching and no one knows the true impact it will have other than it will be felt across the entire search landscape. How fast this change will happen is anyone’s guess but it will be far sooner than anyone can predict and preparation will be a vital component to mitigate disruption and guarantee the best for clients.


Luke Holman - Senior Organic Performance Executive