21 / January 2016, Matt
On October 26, 2015, Google announced through Bloomberg that it had introduced a new signal for its search algorithm: RankBrain. This program represents a step further in the direction toward using artificial intelligence to rank web searches. The system has been in place for a few months now, and is responsible, as Google states it, for interpreting a “very large fraction” of search queries entered into Google.
While humans still play an integral part in Google’s interpretation of searches and ranking of results, RankBrain continues to develop and push Google further in the direction of helping consumers find the information they are looking for, by processing connections between words and their meanings faster than has ever been done before. For those who do SEO the right way on their websites, this should be good news.
When most people think of artificial intelligence, the obvious pop culture reference is Watson, the IBM supercomputer that performed on Jeopardy! in 2011. The technology behind the performance consisted not of thinking, but rather of making incredibly fast conceptual connections among words and phrases. The computer arrives at these connections through a combination of information that people program in and analytics software that uses contextual distinctions to determine what words mean when placed in the order and context they are, as opposed to in another order or situation.
RankBrain does all of this in the realm of search strings. While Google has been doing conceptual searching for several years, this release is a leap forward in those efforts, and a culmination of five years working in the area of artificial intelligence. It had actually been operating for several months before the release. Google has not specified what percentage of searches that RankBrain has processed, but it is significant, and it will continue to grow.
Despite the name, this program does not equate to a human mind. It runs objective analysis of searches and search results, and hones the rankings it produces by doing this over and over again. It does not introduce thought into the process, but what it does accomplish is running through data much faster, and with much more objective accuracy, than an actual mind can do. It examines details with consistency that humans cannot deliver, and in doing so gives Google the ability to push its search capabilities forward much faster than mere manpower could do.
That said, RankBrain does not replace people, but complements them. Humans will continue to load in data, to analyze and measure results, and to work with the system to make it work as efficiently and accurately as possible. Moreover, the goal of RankBrain, and indeed of any artificial intelligence system, will always be to approximate the way a person thinks. It represents the current peak of human ability and performance in its programming, but will continue to benefit every day from that ingenuity in its development.
Conceptual searching is of course not new to Google; it has been working for quite some time to move the significance of keywords away from literal meanings and toward what searchers really want to find. And it has made progress in that over the years. Keyword stuffing without the content to back it up does not achieve nearly the site ranking success it once did, and excellent content that focuses on what users want to read builds a rank much higher than keywords achieve by themselves.
Still, RankBrain does more than previous models achieved. The difference now is the ability of the system to “learn,” in a manner of speaking. Each day, about 15 percent of the searches entered into Google are unique—that is, searches that have never been entered before. These are often long-tail queries, which contain more words and tend to be more complex than shorter queries.
RankBrain has the ability to do two important things with these queries. First, it traces information across vectors to draw connections among seemingly unrelated words, and lead to results that match up to what the searcher wants to find, much more accurately than Google has been able to do before. It interprets the intent of the search and pulls in with greater precision results that lead the user to pages with content that matches that intent.
Second, and perhaps more significantly, RankBrain learns over time. Rather than relying on the learnings of programmers who enter updates into Google’s databases over time, it continues to draw and create connections based on the searches entered and data drawn in over time. All of this learning takes place offline; historical searches and data are loaded into the system, and it runs analytics to make the complex connections that are required to make it go. Thus, to say RankBrain learns as it runs would be a bit misleading. Rather, it continuously receives, processes, and adjusts to data that Google has produced over time to continually improve the results it helps bring.
While it still requires human workers to do quality control and to fine-tune the results, RankBrain represents a big step toward true automation of the machine learning process. The goal remains asymptotic, as humans will always be tinkering, fine-tuning, and improving on what the system can do. Still, the possibilities are exciting as Google hurtles toward the future.
While RankBrain’s functionality may feel magical, it actually comes from rigorous work and processes applied over time. The first step is to provide the system with large volumes of data—and Google has that in spades. With about three billion searches entered and processed every day, and with hundreds of millions of websites in existence, the web of data from which Google can draw is virtually endless. It is limited not by the data available, but only by the speed and accuracy by which it can process.
Getting data is only the first step of the process. From there, RankBrain begins to make things happen. A computer can sort through data more quickly and more accurately than a human can ever hope to do. Now, though, instead of performing mathematics, the machine is performing qualitative and quantitative analysis on words, looking at how words appear, where they appear, and what other words and phrases appear around them. It can determine that “President Obama’s wife” refers to Michele Obama, and that depending on context, “mercury” can refer to a liquid metal, a planet, or a spacecraft. The more it processes, the more long-tail searches it can untangle into meaningful requests for information.
Sorting data in this manner does not become perfect after one iteration. After each “learning,” RankBrain’s results are tested rigorously. A way to think of it is to compare the process to ice sculpting. The first cuts give a sculpture its basic shape, but each pass over the block of ice serves to carve out more details, to hone the sculpture in finer detail until the sculptor creates something incredible. So it is with RankBrain; every pass through the data gives it a better, clearer picture of what the words and phrases and contexts of a search mean.
With so many new, unique searches every day, this process does not have a conclusion. Instead, it continues in different iterations, across different data sets. Every day new information is entered and processed, and every day RankBrain can move a little closer toward “understanding” what people really want to find when they enter a search query.
Google has stated that RankBrain is now the third most important signal in Hummingbird, its overall search algorithm. It has not clarified what the first two are, but given Google’s investment and the speed with which it has come into prominent use, it is safe to say that RankBrain matters a great deal to how searches are processed and pages ranked in response to those searches. As the program develops, one would expect it, or future versions of it, will gain more relevance over time.
Still, RankBrain remains one of more than 200 major signals, that comprise Hummingbird. Further, those signals can be broken down into thousands of subsignals, each of which plays a part in how Google ultimately interprets and ranks searches. Keywords and phrases continue to matter, as will local SEO techniques, page authority rankings, and everything else that lifts one website higher than another in page rankings.
All of these signals work toward same goal: connecting people to information. The current overall effect on SEO is small, but noticeable. Its primary importance lies in how it works, and why. Creating searches is becoming less a matter of precise craftsmanship over time; changes in syntax that don’t alter the meaning of a search make less and less of a difference in the results that a search yields. Moreover, longer search strings are now starting to lead toward more precise results rather than more frequent inconsistencies.
This matters largely because many of the unique searches each day come from those long strings. People don’t think in terms of keywords and short phrases. They don’t train in what SEO phrases others are searching for, so they don’t search on the basis of those. They search based on complex human thoughts and fully-formed questions, or on ideas that don’t fit neatly into a two- or three-word phrase.
The more programs like RankBrain do to accommodate these searches, the better they serve the needs of people who use them. While other signals still matter, and according to Google even matter more than RankBrain, all of them work toward that push to compatibility with how people think about search. Interpreting what people want and connecting them to it is happening more effectively every day.
Equally importantly, the introduction of RankBrain functions as a reward to SEO providers who refrain from gimmicks and deliver consistent, quality writing. When a system is designed to analyze language patterns to understand and interpret information provided, plugging in keyword phrases in the exact order every time and in heavy concentrations becomes less useful. Information that makes sense to a human reader and information that the computer reads and pulls in response to queries are aligning more than ever before. Quality writing matters, both to humans and to Hummingbird.
As RankBrain develops, this will become even more important. Quality SEO writing will bear striking similarities to other quality writing. You can write naturally, with phrases worded differently across your web content where it makes sense to do so. Your flow should not be interrupted by getting in that precisely worded key phrase sprinkled in every hundred words, because RankBrain will understand the concept to which you are writing in a way that matters more than the wordplay that even today interrupts writing quality.
Of course, just because keywords shouldn’t drive the focus of your writing doesn’t mean they don’t matter. Keyword phrases will serve the dual function of sending that additional signal to the computer and helping you focus your writing to a specific topic. Excellent writing sticks to a theme, a particular concept that you want to build around. Failure to stay on point will take away from the authority you present in your page; just as a high school essay needs to stay connected to its thesis, a page of web content should remain structured around a topic. When you accomplish this, the system will reward your content because there is more information on a single area, instead of wandering in different directions over the course of the page without really delivering on any of them.
Instead of spreading your content in too many directions, then, you need to focus on what your core product and/or service are. What do you know, and what can you offer that others cannot? A content campaign that you build around this, approaching that information from different ways and delivering your knowledge and experience in a way that other people can benefit, will serve you very well. RankBrain functions specifically to bring the most useful content to users, so the more you do to provide this, the better it will be for your marketing campaign.
Finally, you cannot rest on the current state of your content, however good it may be. RankBrain will continue to process data every day, and for your site to keep up and maintain its rankings, you need to keep producing high-quality content. Again, this fulfills Google’s goals, because it rewards content that is unique and current. The more you can create what people need, the more your website will benefit from RankBrain. The search for information requires new information to come out, and when you provide that, your campaign will benefit.
At 1st in SEO, we work to provide you with SEO writing that gets results. We have never built anything on the basis of tricking a machine. We want to create content for you that people want to read. After all, optimizing your website should mean more than getting initial clicks from people who accidentally come to your site; it means creating content that brings people in, keeps them engaged, and makes them want to work with your company.
Because this is our approach, RankBrain represents a breakthrough that only reinforces the value that 1st in SEO can provide to you and your company. People are searching for information on companies like yours every day, in a variety of ways, both in local-oriented searches and conceptually-oriented searches. They search for your name, for your product, and for your service level. RankBrain focuses people toward what they want, however they go about it.
To take advantage of this, you need to make clear in your content marketing what you do, and how and why you do it. We work with you to identify your company’s true strengths, and deliver content that focuses its attentions there. We will help you build out a content library with focus and intent, a content library that doesn’t merely attract people to your website, but keeps them there to convert their interest into sales.
Google’s movement points squarely toward rewarding excellent content that gives people what they want. You can benefit from this as long as your approach to marketing is the same as your approach to business: taking care of customers by giving them what they want and need. Contact 1st in SEO today, and we will help get you there.