Post by Erik Bergstrom, SEO Analyst
On October 26th, Bloomberg Business announced that Google is now using a new artificial intelligence (AI) system called “RankBrain.” Not only is Google using a new AI, it’s considered a machine-learning AI system that can teach itself based on what it encounters. Not much information has been released on its inner workings, but from what we know from the Bloomberg article and other SEO outlets, this functionality is meant to help Google’s search algorithm understand more diverse or indirect queries.
WHY IS THIS HAPPENING?
RankBrain has been confirmed as a large part of Google’s search algorithm, Hummingbird. Google has claimed that the search algorithm is the third-most important ranking factor. Google has a history of being purposefully vague about factors that play into its algorithm and how important each piece is to the puzzle. In fact, Google hasn’t outlined what the top two factors above this AI are. So how does RankBrain fit into this puzzle? From what we understand, it’s a way to interpret the intent behind the search even if the right words aren’t used.
This may sound confusing or (or even terrifying), but Google has been able to correctly identify indirect intent behind a majority of searches due to stemming and improvements to identifying synonyms. “Indirect intent” includes long-tail queries that never explicitly reference the word(s) in question, with Google still returning what the searcher was looking for. We can see this exemplified through Google’s Knowledge Graph, which can provide instructions on how to prepare that dish you forgot how to spell or confidently answer a question like: “when was Reagan’s wife born?” Google understands what the searcher is trying to reference, and the Knowledge Graph would be returned with Nancy Reagan’s birthdate without the user directly typing her name.
We know Google has had this ability for some time now, so why create RankBrain? In the past, human labor has been the catalyst to this ability, requiring work to be done on the back-end to acknowledge these implicit connections identifiable by humans, but hard for a machine to learn. RankBrain changes this dynamic, allowing an AI system to learn to identify connections between long-tail queries and what users actually want to see. Beyond making connections, it has also been confirmed by Google that RankBrain does play a role in how well pages will rank. How exactly it factors in has not been stated.
So how does RankBrain’s learning process work? Google told Search Engine Land in their RankBrain FAQ that Google provides batches of historical searches to RankBrain while offline and that predictions are to be made from past searches. After each batch, the latest version of RankBrain is tested and if proven acceptable, it goes live.
CONSIDERATIONS for BRANDS
Google told Bloomberg that RankBrain has been live globally for a couple of months and processes “[a] very large fraction” of queries. Currently, Performics sees no reason to worry about this having a major impact on how well your site is ranking.
For more information on Google’s RankBrain, the new AI system, please contact your Performics Account Team today.