Google Unveils RankBrain Machine Learning For Search

The Google search engine has unveiled a new machine learning system called RankBrain. The new system makes up part of the overall Hummingbird algorithm that displays search results. It was designed to let the search engine handle more complex searches more naturally and understand context, rather than focus solely on the keyword term. It should be noted that, despite the seeming importance of the system, it is one of around 200 or so different factors that are used by Google in its ranking process.


What is Google RankBrain?

RankBrain is mainly used as a way to interpret the searches that people submit to find pages that might not have the exact words that were searched for. This differs from the earlier methods, such as stemming words and using dictionary synonyms to create search results from similar keywords. The knowledge graph also works in tandem with the existing systems.

Within this existing technology, there are around 15% of queries each day that has never been seen before; this is where the new part of the algorithm comes in to play. The system is best suited to answering the ‘long tail’ more conversational query’s that are typed (and presumably voice dictated) into the search.


How Does Google RankBrain Work?

The announcement of the change was, perhaps unusually, given to the business website, rather than being announced through Google’s own PR channels. Whilst notably light on the technical side, the article did give one example of the type of questions that were being passed to RankBrain. The sample search term was:

“What’s the title of the consumer at the highest level of a food chain”

The suspicion amongst SEO’s is that the purpose of the system is to take this long and unwieldy question and translate it to a more commonly asked, and shorter, question that the search engine knows the answer to.

A recent article by Danny Sullivan on the Search Engine Land blog also suggested that that system may also be used to make guesses regarding the correct version of English to use in a result term. For instance, terms such as tablespoon and cup when used in recipes have a different meaning when used in Australia and the UK, compared to their meaning in the US. It is this level of intelligence that it is supposed that the new software is focusing on.


Marketing Using Google RankBrain?

From a marketing point of view, it means that you are likely to see relatively small growth in search results activity around long tail keywords, presuming that you have something relevant to offer the 15% of queries that are in the system.

The update is therefore consistent with Google’s long-held vision of bringing relevant searches to its customers, using the highest quality content it can find. If nothing else this has created yet another reason to take your time to create high quality and informative copy. It is also rather difficult to suggest any particular route to take advantage of this, as the RankBrain tool is largely used when Google itself is unsure of the results to serve.

For most sites though, it will filter into the mix of results, almost imperceptibly in smaller terms.