Google uses a sophisticated artificial intelligence system to help process search results. That system is called RankBrain.
The existence of RankBrain is important to you as a digital marketer because, as the name implies, it affects how Google ranks pages. If you’re at all interested in reaching the top of the search engine results pages (SERPs), it’s in your best interest to know as much as you can about how RankBrain processes search results.
In this article, we’ll cover RankBrain in some detail. Then, we’ll explain how it will affect your search engine optimization (SEO) efforts going forward.
What is RankBrain?
RankBrain is a machine-learning artificial intelligence algorithm that the Google search engine uses to process search results. There’s quite a bit in that sentence so it’s best to break it down into some level of detail.
For starters, the word “algorithm” is a fancy word used by computer geeks that basically means “software code that solves a problem.” In this case, the “problem” that Google is trying to solve is: which one of the countless web pages in cyberspace deserves to be ranked highest for a particular search term?
It’s important to note at this point that RankBrain helps solve that problem. However, RankBrain is not, by itself, the solution to the problem. More on this in a bit.
The phrase “machine learning” means that the software is self-taught. In other words, it isn’t just responding to code written by software developers. It’s learning on its own and adjusting its algorithm accordingly.
The phrase “artificial intelligence” refers to the idea of a computer acting like a human being, but with much faster processing power when it comes to solving mathematical problems. Basically, Google is using computer speed to analyze page rank from a human perspective.
When many people think of artificial intelligence (or “AI” for short), they think of sci-fi movies like The Matrix or Terminator. While quite a bit has been written about computers being self-aware and turning against humanity, the reality is that AI is a great way to automate information systems processing.
In fact, it’s likely that you’ve already seen AI in action on more than one occasion. Facebook uses AI to filter your default newsfeed by showing you items that you’ll find most interesting. Of course, Microsoft has been using AI for a while to facilitate searches on its own Bing search engine.
The difference between machine learning and AI is often debated. For our purposes, machine learning is the ability of the software to acquire information about changes in search results. AI, on the other hand, uses that acquired information to optimize the SERPs.
Finally, RankBrain “processes search results.” That’s different than other algorithms Google uses to determine page rank. As we’ll see, Google tends to look at content and backlinks in addition to a variety of other factors when determining where a pages should end up in the SERPs. The RankBrain algorithm, on the other hand, analyzes searches instead of content. We’ll see how it does that later on.
RankBrain Isn’t the Search Algorithm
RankBrain is only a part of Google’s search algorithm. It isn’t the entire algorithm by itself.
Many people are under the impression that, because news broke about RankBrain recently, it’s a change to the overall search algorithm. That’s not the case.
Instead, RankBrain is just one small part of the much larger digital ecosystem that makes up the Google search algorithm.
For a long time, the Google search algorithm didn’t have a name. Around the middle of 2013, Google officially dubbed it “Hummingbird.”
RankBrain is just one part of Hummingbird, but there are other parts as well. Other Hummingbird components that are familiar (and, sometimes, infamous) to SEO professionals include the Panda and Penguin algorithms, used to fight link spam. Another component, called Top Heavy, is used to prevent pages with too many ads near the top from rising high in the SERPs. There’s also the Mobile-Friendly component that lives up to its name by promoting sites that offer quality content for mobile users.
Those are all parts of the Hummingbird search engine algorithm that are separate from RankBrain.
What About PageRank?
If you’re a savvy SEO professional, you might have heard about a Google algorithm called PageRank. At this point, you might be asking yourself: “What’s the difference between PageRank and RankBrain?”
PageRank is a separate part of Hummingbird. It’s used to give credit to pages with quality backlinks. If you’ve ever tried to get links to your website (or your client’s website) from pages like The Huffington Post or Forbes, then you’re trying to use the PageRank algorithm for your own benefit. Google has been using that algorithm since 1998.
Google also uses “signals” to determine page rank. Those are various indicators (e.g., bold words, mobile-friendliness) that Google uses to determine quality and substance. PageRank will process that information as well and use it to determine where the page appears in the search results for a given keyword.
However, there are more than 200 ranking signals Google uses to determine placement. Beyond that, there are 10,000 so-called sub-signals or variations on the signals themselves. Important to note, Google is killing off the idea of PageRank, but they will still be using the concept internally to some degree.
That brings us back to RankBrain.
RankBrain is a Signal
Now that you know something about the Hummingbird algorithm and how it processes search results, it’s time to learn a little bit more about RankBrain.
It’s this simple: RankBrain is one of the signals that Google uses to determine rank.
As we’ve seen, though, it’s not the only signal. So it’s important to consider what RankBrain does as part of the larger picture about how Google determines page rank.
According to Google, RankBrain is the third most important signal when it comes to ranking web pages. Here’s what Bloomberg had to say on the subject:
RankBrain is one of the “hundreds” of signals that go into an algorithm that determines what results appear on a Google search page and where they are ranked, [Greg Corrado, Google senior research scientist] said. In the few months it has been deployed, RankBrain has become the third-most important signal contributing to the result of a search query, he said.
So right now you might be asking yourself: “If RankBrain is the third-most important signal when it comes to determining page rank, what are the first two?”
That’s an excellent question.
A Reluctant Admission
At first, Google wouldn’t disclose any information about the first two most important signals. Even after repeated queries from well-respected members of SEO media, the folks at Google refused to spill the beans.
That led many people to ask why Google announced the third most important ranking factor without telling anybody about the first two. The answer was simple: to get more press.
In all likelihood, Google wanted to brag about its new algorithm. The emergence of RankBrain was a good way for the company to keep its name in the spotlight and highlight a major milestone.
Finally, though, Google realized the error of its ways and acknowledged that the first two ranking factors are exactly what many SEO practitioners had suspected all along— backlinks and content.
Those are in no particular order, though.
Backlinks are links to a particular web page from another website. It’s implied, though not specifically expressed, that backlinks from high-profile sites like Mashable or Business Insider offer better rank signals than backlinks from lower quality blogs.
The content signal is based on what’s actually written on the website. If you write a high quality article about a unique subject like insomnia on an airplane, you can be fairly certain that it will rank well when people search for “insomnia on an airplane.” That’s especially true if you have a blog that regularly produces content that Google has already ranked.
Those signals are different from RankBrain, though.
What is the RankBrain Signal?
RankBrain goes beyond backlinks and content to find pages that should rank high for a specific query based on factors that other signals might have missed.
For example, some keywords or phrases that people search for might not appear on a specific page even though that page is packed with lots of information relevant to the particular search. Certainly, Google has been able to find pages that don’t contain keywords for many years. In those cases, though, the search algorithm uses synonyms (e.g., “footwear” for “shoe”) or stemming (e.g., “dresses” for “dress”).
Google also uses a Knowledge Graph to create smarter connections between key words or phrases. In that case, Google searches for “things not strings” (the word “strings” here being IT speak for words or phrases).
The knowledge graph views words as not just a collection of letters that it’s looking to match on web pages in cyberspace, but it views them as real-world entities. For example, if you’re searching for “Hillary” right around now, you’re probably looking for the former First Lady who’s running for president. Sure enough, if you Google “Hillary,” you’ll find that the first few results, as of this writing, are all about Hillary Clinton. That’s the case even though there are countless people named Hillary.
The Knowledge Graph is Google’s database that maintains information about how things in the world are related to one another. That’s why you can type “When is Labor Day” in the Google search bar and immediately be presented with an answer about the exact date of Labor Day this year.
Even though Google does all of that, though, RankBrain is still necessary.
Google currently processes roughly 3.5 billion searches per day. If you break this statistic down further, it means that Google processes more than 40,000 search queries per second.
Of those searches, 500 million search queries—which comprise 15% of all queries submitted on Google—have never been seen before by Google’s algorithms.
Let that sink in for a moment. Those unique queries represent a search for information that nobody has performed before.
Many of those queries are so-called “long-tail” searches. That’s when the user doesn’t just type a short phrase (e.g., “discount laptops”) but a lengthy, very descriptive search phrase (e.g., “refurbished late model Dell laptops”).
RankBrain exists to parse and interpret exactly those types of queries so as to provide a better, more useful series of search results to the user. The algorithm detects patterns between complex searches that are apparently unconnected and employs its “machine learning” to discern how they’re actually similar to each other. Then, it uses AI to produce results based on what it’s learned from comparing the searches. RankBrain also associates groups of searches with results that will offer a better user experience.
In a nutshell, RankBrain will translate complex or ambiguous searches into something more specific based on existing search data. That way, it can offer page results that will most likely benefit the user.
Here’s more about RankBrain from Bloomberg:
RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand. 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.
The RankBrain Signaling System
How, exactly, does RankBrain work as a signal to factor into page rank?
Again, Google isn’t offering too many details about the algorithm. What we do know is that it uses query refinement to generate signals that determine rank.
While it’s historically been the case that signals are tied to content, RankBrain signals are generated as a result of the search queries themselves. That much has been confirmed by Google.
How does the signal get created? It’s probably using the queries and the existing results to create a better summary of a page than Google’s existing algorithms.
Query Example 1
Fortunately, Google was kind enough to offer an example of how RankBrain parses and processes search queries.
Consider the following query: “What’s the title of the consumer at the highest level of a food chain?”
Think about how that query could easily be misinterpreted by a computer processing it as a series of strings. The phrase “food chain,” for example, could refer to a national chain of grocery stores. Indeed, even an intelligent system might think that the query has something to do with a retail environment because the word “consumer” is also in it.
So what’s the right answer to the query? A predator.
Read it again and you’ll see that it’s really asking who’s at the top of the food chain. The phrase “title of the consumer” is just an awkward way of phrasing the question.
Right now, if you enter that query into Google, you’ll get results that indicate Google is answering the question the way that it should be answered. The first organic entry is a Wikipedia page about food chains. The next entry has most of the search phrase in the article and it’s also about predators, so that’s a good result as well.
Now if you Google “top level of the food chain,” you’ll see that you get similar results.
What’s RankBrain doing here? While details about the algorithm are hush-hush, it looks like it’s connecting the very long and awkwardly worded query to the second, more common-sense one. It’s using “machine learning” to determine that the two queries are very similar and adding that information to its portfolio of knowledge. RankBrain’s AI will then use that knowledge in the future to provide users with a better set of search results.
Keep in mind that the example we’ve just looked at is one offered by Google, so it’s safe to say that it’s a valid demonstration of how RankBrain “thinks.”
Query Example 2
The folks at Google also provided Search Engine Land with an additional example so we can all get a glimpse of the power of RankBrain.
Consider this query: “How many tablespoons in a cup?”
That might seem like a question with an easy answer. It just involves a little bit of parsing and some math. It shouldn’t be any problem, right?
The answer to that question varies by country. If you’re in Australia, for example, the answer will be different than if you’re in the United States.
According to Google, RankBrain pushes certain pages higher in the SERPs depending on the country of origin where the query occurred. That means people who issue that query in Australia will see results different than those who issue the same query in the United States. More importantly, in each case the results will be better suited to the user.
That’s according to Google. However, queries run right now using Google.com and Google.com.au yield the same results. Of course, it’s possible that Google is smart enough to know the user’s country even when that user is visiting a foreign Google site. So we’ll just have to take Google’s word for it in this case, unless you know somebody in Australia that you can ping.
RankBrain is also relatively new. It might be the case that the algorithm hasn’t quite “matured” enough yet to process that difference, but will one day get to that point.
When Did RankBrain Go Live?
The new algorithm started with a gradual rollout in early 2015, after Senior Vice President of Search Amit Singhal (who is not at Google anymore) greenlighted the project. It went fully live worldwide just a few months ago.
It wasn’t easy to get to that point, though. The official rollout was the result of a yearlong effort of a small group of Google engineers. That team included search specialist Yonghui Wu and deep-learning expert Tom Strohmann.
How Many Queries Does RankBrain Handle?
According to Google, RankBrain currently processes a “very large fraction” of all Google queries. That’s as much detail as Google is willing to give at this point.
Of course, a “very large fraction” could be a majority or a minority.
The Proof is in the Pudding
How does Google view these first few months of a fully deployed RankBrain? According to the company, RankBrain has been a resounding success.
“I was surprised,” said Greg Corrado, a senior research scientist with Google. “I would describe this as having gone better than we would have expected.”
CEO Sundar Pichai concurred, signaling on a recent conference call that the algorithm is symbolic of the strategic direction of the company.
“Machine learning is a core transformative way by which we are rethinking everything we are doing,” he said.
It also appears that RankBrain is outthinking humans, as it should. Recently, Google search engineers were asked to look at some pages and guess which ones they thought Google would rank on top. They were right 70% of the time, but RankBrain went through the same test and was right 80% of the time.
Corrado said that turning off RankBrain now “would be as damaging to users as forgetting to serve half the pages on Wikipedia.”
“Search is the cornerstone of Google,” he said. “Machine learning isn’t just a magic syrup that you pour onto a problem and it makes it better. It took a lot of thought and care in order to build something that we really thought was worth doing.”
What About Bing’s RankNet?
As a good digital marketer, you already know that you shouldn’t just concern yourself with Google, even though it’s the undisputed king of search engines. Bing is also used by many people for online searches.
Bing’s machine-learning system is called RankNet. It’s part of the modern-day Bing search engine.
Although Microsoft hasn’t said much about RankNet, it appears to function very similarly to RankBrain. In fact, it’s possible that Google got the idea for RankBrain from Bing’s RankNet.
In fact, if you plug the query from above (“What’s the title of the consumer at the highest level of a food chain?”) into the Bing search engine query bar, you’ll find that you get similar results. As of this writing, the result that features the phrase almost identical to the search phrase is on top, whereas the Wikipedia entry is in second position.
If you’d like to learn more about how Google captures similarity between concepts, the company has posted a bit about the subject on its open source blog. Here’s how Google describes it:
We’ve shown that computers can learn to recognize cats (and many other objects) just by observing large amount of images, without being trained explicitly on what a cat looks like. Now we apply neural networks to understanding words by having them “read” vast quantities of text on the web. We’re scaling this approach to datasets thousands of times larger than what has been possible before, and we’ve seen a dramatic improvement of performance — but we think it could be even better.
If you’re a propeller-head who likes to tinker with software development on your own, you’ll be happy to know that Google also provides its own open source software “for computing vector representations of words.”
If you’re interested in a more formal, academic approach to the subject, you can check out this research paper.
What All of This Means for SEO
At this point, you should have as much knowledge about RankBrain as is publicly available. Now, you might be wondering how you should alter your SEO efforts.
The short answer is: probably not much.
The old-fashioned practices of getting backlinks from solid websites and producing outstanding content that doesn’t overstuff keywords will still help you rank sites. You should continue those practices for now.
Since search intent has become a priority, RankBrain only gives preference to the pages that truly meet its requirements. Basically, optimizing content is not possible without understanding the search intent and performing intent-specific keyword research.
With that being said, optimizing for long-tail keywords could potentially have an adverse affect on your site’s overall user experience.
Instead, aim for medium-tail keywords, that aren’t as competitive as the top-tier keywords and still have a bit of search volume. We recommend building your page around these medium-tail keywords.
Frequently Asked Questions
1. What is Google RankBrain?
RankBrain is a machine-learning system that Google uses to interpret the user intent of a search query and provide those users with more relevant search results. Originally rolled out in 2015, RankBrain has become Google’s third most important ranking signal.
2. What is RankBrain used for?
RankBrain is used to process searches and refine queries. In order to accurately gauge a searcher’s intent, Google feeds RankBrain a substantial amount of data. Then, RankBrain analyzes that data and teaches itself how to serve up the most relevant information based on certain search signals, like search history, device, and location.
3. Why is RankBrain important?
RankBrain is fundamentally important to the future of search. Not only does it help relatively rare and ambiguous queries get translated to more concise and relevant terms, but Google recently indicated that this part of the algorithm is now a critical component of organic search and the user experience.
RankBrain is the newest piece of the search engine algorithm that Google uses to rank pages. It’s also the third most important ranking signal.
Undoubtedly, RankBrain will play an important and, in all likelihood, an increasingly important part of determining page rank for years to come.