Launch date: October 26, 2015 Hazards: Lack of query-specific relevance features; shallow content; poor UXHow it works: RankBrain is part of Google’s Hummingbird algorithm. It is a machine learning system that helps Google understand the meaning behind queries, and serve best-matching search results in response to those queries
What is Google RankBrain?
DATE GOOGLE CONFIRMED EXISTENCE OF RANKBRAIN: OCTOBER 26TH, 2015
RankBrain is a component of Google’s core algorithm which uses machine learning (the ability of machines to teach themselves from data inputs) to determine the most relevant results to search engine queries. Pre-RankBrain, Google utilized its basic algorithm to determine which results to show for a given query. Post-RankBrain, it is believed that the query now goes through an interpretation model that can apply possible factors like the location of the searcher, personalization, and the words of the query to determine the searcher’s true intent. By discerning this true intent, Google can deliver more relevant results.
The machine learning aspect of RankBrain is what sets it apart from other updates. To “teach” the RankBrain algorithm to produce useful search results, Google first “feeds” it data from a variety of sources. The algorithm then takes it from there, calculating and teaching itself over time to match a variety of signals to a variety of results and to order search engine rankings based on these calculations.
To clearly conceptualize RankBrain, it can help to put yourself in Google’s shoes, trying to understand the intent of a search engine query like “Olympics location.”
What is the true intent of this search? Does the searcher want to know about the Summer or Winter Olympic Games? Are they referring to an Olympics that just concluded or one that will take place four years from now? Is the searcher attending the Olympics right now, sitting in a hotel and looking for directions to the venue for the opening ceremonies? Could they even be looking for historic information about the location of the very first Olympics in ancient Greece?
Now, imagine that in trying to answer this query, all you have is simplistic algorithm signals like the quality of content or the number of links a piece of content has earned to rank results for this searcher. Imagine that the Winter Games in Sochi, Russia just concluded last month and the official Sochi Olympics website has earned millions of links for its content about this past event. If your algorithm is simplistic, it may only show results about the Sochi Games, because they have earned the most links… even if the searcher was actually hoping to learn the location of the next Winter Olympics in Pyeongchang, South Korea.
It’s within this complicated but common situation that the capacity of RankBrain emerges as essential. It’s only by being able to mathematically calculate results based on patterns the machine learning algorithm has “noticed” in searcher behavior that Google can determine that, for example, the majority of people looking up “Olympics location” want to know where the very next Games (be they Summer or Winter) will be held. So, in this case, a Google answer box with the upcoming Games’ location in it will serve the majority of searchers’ needs.
While that answer box may address the intent behind most “Olympics location” searches, there are notable exceptions Google must address. For instance, if the search is being performed by a user within an Olympic city (like Pyeongchang) the week of the games, Google might instead provide driving directions to the pavilion where the opening ceremonies will be held. In other words, signals like user location and content freshness must be taken into account to interpret intent and deliver the results most likely to satisfy searchers.
*RankBrain is a work in progress, with the goal of machine learning perfecting Google’s interpretation of searcher intent over time. Interestingly, our hypothetical query, “Olympics location,” performed in the United States in April of 2017 is returning this Google answer box result:
Does this indicate that the machine believes most people searching for this term are still more interested in the 2016 Rio de Janeiro Summer Games than they are in the next event, the 2018 Pyeongchang Winter Games? Is RankBrain succeeding here, based on patterns it has calculated or is it still “in the works,” not sure from the vagueness of our query whether we want an older, popular answer, or a fresher one that looks to the future? And what would this query return if we could perform it in January of 2018? Would the answer box show Pyeongchang because the signals surrounding the event have intensified by that time?
Because of the extent and nuances of RankBrain’s influence on how Google’s core search algorithm works haven’t yet been fully fleshed out, one of the best ways of learning more about how RankBrain works may come from observing how often Google responds to a variety of your own queries with satisfying answers. How often to they correctly interpret your intent?
Does RankBrain change the way we do SEO?
Depending on the sophistication and modernity of your personal SEO skills, RankBrain may represent either a minor or major change in your theories and practices. Respected patent expert Bill Slawski provided the following illustrative example of why RankBrain is necessary in the search environment:
“To an equestrian a horse is a large 4 legged animal, to a carpenter, a horse has 4 legs, but it doesn’t live in fields or chew hay, to a gymnast a horse is something I believe you do vaults upon; with RankBrain context matters, and making sure you capture that context is possibly a key to optimizing for this machine learning approach.”
1. Different rankings signals apply to different queries
Pre-RankBrain, it might have been appropriate to assess website page optimization by evaluating all traditional signals (link diversity, content depth, keyword matching, etc.). Post-RankBrain, SEOs need to determine the type of content that best serves users’ needs. For something like a sudden hurricane, you’re going to count on freshness much more than the links a piece might have accrued. For something like the history of Indigenous American music, you’ll be relying on content depth, and possibly related topics your domain covers, signaling authority. Know that the machine learning algorithms that drive RankBrain are matching signals to query intent and that SEOs must do this, too.
2. Signals apply to your website’s reputation
SEO seeks to build your brand’s reputation as a resource trusted by search engines and human users for providing a specific experience. The benefits of establishing such a reputation can include ranking well for the keywords most important to you. Does your brand need to build its reputation on freshness, depth, diversity of earned links, high user engagement, or other signals? The answer depends on the topics you cover (e.g., real-time sporting events scores vs. an online course in learning the Spanish language). Do the searches you hope to rank for demand quick, brief answers or in-depth explorations? Over time, your domain must build a reputation based on the signals it wants to serve, realizing that RankBrain creates an environment in which your brand can become known for delivering a particular type of content that satisfies a particular need.
3. One-keyword-one-page is really, really dead
Likely, you already know that the practice of creating a page for “spatula,” another for “spatulas,” another for “kitchen spatula,” another for “pancake turner,” and another for “metal spatula” is a tired old horse that needs to be put out to pasture. Modern SEO would combine all of these phrases (and their associated URLs) into a single piece of thorough content that incorporates natural language, including variant keyword phrases that reflect the way humans search and speak. This is not new news for most alert SEOs, but the advent of RankBrain highlights the wisdom of focusing on total keyword concepts with comprehensive content, rather than breaking out multiple pages to cover variants like “widget” vs. “widgets.”
Other facts about RankBrain
RankBrain has been cited as part of the overall Google Hummingbird algorithm.
In 2015, Google stated that RankBrain was being used to process 15% of queries the system had never encountered before. By 2016, Google was applying RankBrain to all queries.