Google recently announced it is now using “RankBrain,” an artificial intelligence system, to enhance the ways it understands searchers’ intent. With “RankBrain,” Google has not only introduced machine learning to its overall search algorithm, but also eliminated the need for its engineers to manually update synonym lists and databases to make connections between searches and intent. Not long ago, Google said it missed about 15 percent of daily searchesbecause of this manual process. That translates into about 450 million searches per day that never get seen by a human. Google’s solution? RankBrain.
How it Works
RankBrain is the latest of many ranking factors in Google’s algorithm, known as Hummingbird. This is consistent with Google’s approach to enhancing — as opposed to replacing — its algorithms.
As Search Engine Land explains, “Hummingbird is the overall search algorithm, just like a car has an overall engine in it. The engine itself may be made up of various parts, such as an oil filter, a fuel pump, a radiator and so on. In the same way, Hummingbird encompasses various parts, with RankBrain being one of the newest.”
To get the scoop on oh how RankBrain was created, Bloomberg spoke with Greg Corrado, a senior research scientist with Google.
“RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities – called vectors – that the computer can understand,” Corrado said. “If RankBrain sees a word or phrase it isn’t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.”
RankBrain in Action
Search Engine Land provides a great example of what RankBrain can do. Using the query “What’s the title of the consumer at the highest level of a food chain,” you’ll see from the screenshot below Google populated sites that knew the term “consumer” in the scientific world is someone/something that consumes food.
When you search for a more condensed version of the above search “top level of the food chain,” you get very similar results:
While it’s debatable if these two searches are correlated based on RankBrain, it demonstrates how Google can utilize user queries to better understand intent within the billions of searches each day.
After a slow roll-out at the beginning of 2015, RankBrain is now fully live and global, and according to Google now the third most important ranking factor in their algorithm.
To make sure RankBrain understands as much as it can, Google told Search Engine Land RankBrain’s learning is done offline. Google sends batches of historical searches to the RankBrain database where it learns to make more predictions. Once these predictions are tested and validated, the newest version of RankBrain goes live, and the process repeats.
As anyone familiar with SEO knows, at the end of the day computers assess our clients’ sites and make decisions based on what they think is the most relevant choice for any particular search. However, there are instances and edge cases where search results are not always optimal, or things the algorithm should be able to filter out still appear. By incorporating machine learning, Google should be able to scale at an accelerated rate and minimize anomalies in search results. This will likely make it increasingly difficult to engage in black hat SEO techniques.
The majority of users now have access to the Internet in their pockets — making it that much easier to rely on Google and other search engines to answer their questions. Google foresaw the importance of anticipating answers to the many questions we typically have each day, and made huge strides with semantic search via their Knowledge Graph.
Google’s Quick Answers database more than likely (although not confirmed) feeds RankBrain the information it needs to understand search queries better. The pages Google selects for Quick Answers are of high authority because of quality, well-structured content. This content is theme-relevant, optimized for a great user experience and designed to answer specific questions closely matching the query. The addition of RankBrain should help Google future-proof itself against platforms such as Siri where interstitial programs perform the actual searches as opposed to the end user.
Ultimately RankBrain should help fulfill Google’s mission to organize the world’s information in ways that are most useful to its audience. This will benefit marketers who base their content strategies on the behavioral needs of their targeted audiences. RankBrain reinforces the need for a clearly defined set of audience segments from which behavioral insights become actionable. This includes:
- Selecting interesting topics for your users
- Creating content that aligns with what those users are consuming
- Structuring the page with user experience in mind
- Implementing white hat SEO practices which align with Google’s Guidelines
- Identifying the right content format for the right audience (video, PDF, mobile, etc.)
- Providing how-to lists when quick-digestible content is needed
- Measuring the impact of all the above
Machine learning technology like RankBrain isn’t new. Microsoft has had its machine learning technology RankNet since 2005. The best any site owner can do to help feed these programs the correct information is to keep in line with SEO best practices, and always keep in mind that valuable content that connects with your customers’ needs and questions will help get you were you need to be.