RankBrain has been one of the hottest topics in the SEO world for months now.
The truth is: we don’t know a lot about it, since the only official source out there is the Bloomberg article on how Google is turning its search engine over to AI.
However, from our own experience and research, we’ve been able to make some educated guesses about how it works.
In order to understand RankBrain, you first need to understand two things:
- Artificial Intelligence (AI) and
- Machine Learning.
and how they apply to Google and its algorithms.
Artificial intelligence is the capability of a computer to gather knowledge by learning, adjusting, and adapting over time.. More advanced AI is capable of building solutions based on what the system already knows from previous experience.
It does this through machine learning.
Machine learning is an approach to building AI.
All systems using machine learning consist of three parts:
- The model
- The parameters
- The learner system.
The model is a system that makes predictions and identifies patterns. It is the only part of the system that a programmer has to create initially. It’s the initial set of instructions that the machine will follow to begin its learning.
Parameters are signals that the learner system uses in order to make modifications to the entire system.
Finally, the learner system adjusts the parameters and the model based on the outcome of each iteration of learning. Put simply, it’s the part of the system that’s responsible for making tweaks and changes to optimize the model and the parameters.
If you want to know more about machine learning, check out this article by Marketing Land.
What Is RankBrain Then?
RankBrain is the artificial intelligence system that Google is now using to handle more than 15% of total queries. It was built using machine learning and has been deployed worldwide.
Check out this short video to see some quick facts about RankBrain:
According to Greg Corrado, a senior research scientist at Google, RankBrain is the third-most important ranking signal.
The main problem RankBrain solves is the difficulty Google has had understanding new (never-before-typed) search queries, which make up 15% of daily searches.
One reason for this is that, since people are using mobile devices more and more, it’s becoming more common to enter search queries through voice rather than the keyboard.
This makes queries harder for Google to understand. It’s much easier for the machine to understand “Bruce Springsteen tickets” than it is to understand “Buy tickets for the Boss.”
RankBrain helps Google understand what words mean in the context of the user’s query. It identifies the meaning of unfamiliar words and phrases that were never before typed into the search box.
And it does so by learning those patterns on its own.
What Differentiates RankBrain from Other Algorithms?
One major thing that’s different about RankBrain is that it doesn’t need any external data from people to optimize itself. Google Panda also uses machine learning but it needs external data, specifically information about website quality, from Google quality raters.
Panda needs humans to feed data to it so it can adjust itself based on the new data.
RankBrain is a deep-learning system. It doesn’t need any external data from humans. It’s capable of learning on its own.
What We Believe RankBrain Is Doing
1. Direct Impact on Ranking Pages
We’re not 100% certain but we suspect RankBrain does much more than query refinement.
We believe it’s crawling websites to measure user experience and adjust rankings for sites (and their relevant keywords) based on the quality of the user experience on those sites. One way it does this is by analyzing micro-conversions (which you can read more about here).
2. Inferring Meaning from Sentences
RankBrain actually enables Google to read sentences and infer meaning from them.
This system is much more advanced than the previous Latent Semantic Indexing (LSI) system, which could only decide whether the topic was relevant to the query based on the occurrence of similar words and phrases to those found on other top pages in the SERPs.
For example, if your client has a website featuring movie reviews and there is a review of a movie that had a sentence like, “That movie was awesome,” RankBrain could help Google understand that phrases like “awesome movie” are referring to movie reviews.
And RankBrain is even more advanced than that. By learning on its own all the time, it can detect the meaning of an entire article, deciding whether a review has a positive or negative tone.
3. It Could Affect Paid Results
Since its main goal is to provide users with more refined results, we think that paid results included in query results are also being targeted and refined by RankBrain.
If the paid results are more targeted, advertisers value the platform more, and are ready to spend more money. It’s a win-win situation for Google and the advertisers.
What’s Documented About RankBrain?
The only resource we have about RankBrain is the Bloomberg article – and it’s pretty vague.
However, we do know some things for sure:
1. It’s An Integral Part of the Overall Algorithm
RankBrain is not a completely different algorithm that stands on its own (like Penguin or Panda). It’s an integral part of Hummingbird, which is the name for Google’s main search algorithm.
2. It Substitutes Terms in Query Rewriting
Bill Slawski sees the roots of RankBrain in a patent that was granted in August 2015.
That patent describes the process of substituting terms in query rewriting for better results.
For more technical details, you can read about it here. The main takeaway is that the patent shows how the machine learning system can substitute certain terms for others in context of other nearby terms.
What Is RankBrain’s Future?
First of all, it’s important to understand that Google is now Alphabet. This means search is not the only business the owners are interested in anymore.
In 2012, Google hired Ray Kurzweil, a world-renowned futurist. His mission was to build a “cybernetic friend” who would know its users better then they know themselves.
In 2013, Google established Calico LLC, a company that researches how to improve the longevity of life.
In 2014, Alphabet acquired DeepMind, a company that specialized in creating learning systems. DeepMind was acquired for $500 million, which means they were serious about machine learning development even then.
As you can see, Google is investing heavily into artificial intelligence and machine learning. The future of search is bright – it won’t be just about links and content anymore: it will be a dynamic field where manipulating the system will be almost impossible.
What Your Business Can Do Today To Prepare for the Future
Don’t Focus Your Services on Individual Keywords
It’s all too easy to focus on ranking individual keywords for your client. But this is not a long-term strategy anymore.
Instead, you should start thinking about how humans type and how they say something into the search box. Then, you create your search strategy around that.
No longer focus on short fixed keywords like “SEO NJ.” Instead, focus on long-tail keywords like “I need SEO help in New Jersey.” This will help get more traffic from the long-tail keywords that make up 15% of daily searches that people are now typing, based on how they would naturally talk.
Focus on Micro-conversions and Interactivity
We know that Google is already looking at user metrics for your clients’ sites. It’s probably the second-most important signal after links.
But the age of links as the main ranking signal is coming to its end. At their core, they’re easy to manipulate.
As we’ve mentioned before, micro-conversions are the new thing. Google is looking more and more at how interactive your site is and how fast the user can achieve their goal.
Make your site user-friendly. While it was previously okay to focus on minor things like keyword density, new systems like RankBrain are making these old tricks obsolete.
Don’t Focus on Domain Authority Too Much
Artificial intelligence is going to change many things.
One of the signals that’s pretty strong right now is domain authority. It gives you a huge advantage over competition, but it’s easy to manipulate by spammers.
We suspect that there’s an AI bot out there (which is part of RankBrain) that’s pre-testing pages to make decisions on which source is the best for a particular query. In the future, every query could end up with different results based on hundreds of factors that will be happening at the same time.
This will gradually decrease the importance of domain authority as content quality will begin to drive results.
Artificial intelligence is already here. You should prepare your business by beginning to phase out old “SEO tricks” and focusing on a high-quality user experience, micro conversions, and an SEO strategy focused on long-tail keywords. Even if some of those tricks still work today, they may become useless tomorrow and even potentially harm website rankings.
The good news is you don’t have to change your strategy too much. You just need to add to it. Great content is always important and natural links – although losing their power day by day – will still play a major role in the overall algorithm.