RankBrain is the new kid on the block and everyone wants to know all about him. Problem is he’s not much of a talker but we do know a few things about RankBrain, how it works and how it learns. We’re going to share with you everything we know and tell you how to optimise for RankBrain.
TL;DR? No problem, just download the PDF to read later.
- What is RankBrain?
- How does RankBrain work and what does it do?
- Does RankBrain actually work?
- How does RankBrain learn?
- How RankBrain makes Google better
- RankBrain answers query intent
- RankBrain understands term association
- RankBrain provides better results for long-tail phrases
- How to optimise for RankBrain
- How RankBrain affects SEO
- How to optimise content for RankBrain
- How to optimise keywords for RankBrain
- Download PDF
RankBrain is Google’s new machine-learning artificial intelligence system for helping process search results. It’s part of Google’s search algorithm, Hummingbird, used to sort through billions of search queries and find the pages it deems most relevant to display to users. Google has said in the past that they have over 200 major ranking signals and have up to 10,000 sub-signals.
So what makes RankBrain so special if there’s over 200 major ranking signals? RankBrain is now Google’s third most important ranking factor. The first and second most important signals have since been revealed by Google as links and content. No major surprises there.
How does RankBrain work and what does it do?
To put it simply, RankBrain is an artificial intelligence system that helps Google’s current search algorithms. It provides better results to user queries by using mathematical processes and an advanced understanding of language semantics to learn more about how people are searching and what they’re searching for. Then RankBrain applies what it’s learnt to future search results. Meaning, while we may not have seen a huge difference in our search results since RankBrain’s rollout, it’s going to continue to get smarter and better at what it does. According to Google, RankBrain is going to make the web an easier and better place to search.
“I was surprised… I would describe this as having gone better than we would have expected.” – Greg Corrado, Senior Research Scientist at Google
Basically, RankBrain is like a much needed “patch” for Hummingbird, Google’s search algorithm, as the too ambiguous search terms and those based on colloquialisms were too difficult for it to understand and provide good results for. RankBrain is the “fix” for this problem and while it’s already going well, as it is continually and automatically learning, it’s only going to get better.
To put it in laymen’s terms, RankBrain is Google’s way of trying to better match user intent behind search queries.
Does RankBrain actually work?
It’s already working. A study by Stone Temple Consulting on whether RankBrain actually improves search results found that 54.6% of hard to understand queries in their sample group showed improvement after RankBrain. And this number is only going to get better as RankBrain continues to learn the intent behind our search queries.
RankBrain is incredibly smart. It turns word-based search queries into number patterns it can understand, meaning that when Google doesn’t automatically already understand the intent behind a user’s search query (which is around 15% of the time), RankBrain can then decipher what a searcher is asking based upon its learnings. And we know it works because it was tested against Google’s engineers. RankBrain outperformed its makers when asked to estimate the relative ranking of a handful of random web pages.
How does RankBrain learn?
RankBrain surprisingly enough, learns offline. It is fed batches of historical searches and learns to make predictions based off them. These predictions are then tested and if proven good, the latest version of RankBrain is rolled out. This learn-offline, test and go live cycle is repeated over and over again. This is how RankBrain works and how it’s only going to make Google a better, more accurate place to search.
The reason this method is so clever, is because it requires little human input to make it work and learn.
Traditionally, Google has had people create stemming, synonym lists etc. to enable more refined searches, requiring a huge amount of human effort. However, RankBrain trains itself with the historical search data that is fed to it to provide it with a constant learning process, in order to continue to refine and optimise. It’s been speculated that in the future RankBrain will have a deep learning process rather than a machine learning one, meaning it won’t need any human input at all. In future all it may need is a dataset, over which it will apply its deep learning process to refine the algorithm.
With over 40,000 search queries per second on average, translating to 3.5 billion searches per day and 1.2 trillion searches per year worldwide, Google’s RankBrain has a lot of data to learn from.
RankBrain is making Google a better place to search by constantly learning from its users and providing better results for given searches by better matching results to a user’s intent, through term association and long-tail phrases. This means that when you type a fairly ambiguous or colloquial query into Google, it’s going to give you a much better result and these results are going to continue to improve as RankBrain learns.
RankBrain answers query intent
RankBrain means that Google is getting better at answering the intent behind a user’s search query. Google gave us the example that if a user asked “how many tablespoons in a cup?” RankBrain would favour different results in Australia as opposed to the United States as the measurements in each country are different, despite similar names. Favouring a country’s own results in a search is nothing new for Google but the fact that RankBrain knows there needs to be differences in results based on the term (not the country) is pretty impressive.
However, the biggest improvement is in RankBrain’s analysis of difficult to understand queries. For example if you were to ask Google “Who’s that actor kid with the crazy eyebrows”. RankBrain knows exactly who you’re asking about and will give you results for Will Poulter.
RankBrain understands term association
RankBrain is helping Google better understand what a user is actually searching for and is getting better at term association. It understands synonyms and groups these search terms. It also understands homographs.
In the same study by Stone Temple Consulting, it was found that RankBrain is giving Google a much better understanding of search terms. The phrase “why are pdfs so weak” returned results in July 2015 for PDF’s of “Why the Iraqi Resistance to the Coalition Invasion Was So Weak”. Definitely not the intent behind the user’s query. But in January 2016, the same query returned very different results, showing that Google now understands that the term ‘weak’ can be used in the context of security encryption.
Source: Stone Temple Consulting
Similarly, Google now understands that if I was to type in “the wire” I could be asking about the tv show, not an actual wire. Before RankBrain this query would have required a manual exception rule to address. RankBrain is able to see the relationships for this term automatically without requiring any kind of manual adjustment. It can do this by making the observations that these words are often capitalised and that the phrase is quite often used with words like tv, episode, season, series, etc.
RankBrain provides better results for long-tail phrases
RankBrain is able to identify patterns between words and understand how they are related to one another. This enables it to refine future searches related to existing topics in order to give more accurate results to users. RankBrain does this by connecting the original, ambiguous query to shorter answers that are relevant. This means that Google is better able to produce answers to even the most unusual queries.
The example given in the original article that broke the news was this rather complex and vague query: “What’s the title of a consumer at the highest level of a food chain”. This ambiguous query is then matched to a more common and shorter query, perhaps something like “top level of the food chain” or something similar.
So should you be optimising for RankBrain? Maybe. But Google isn’t giving away any information other than RankBrain is in fact a ranking signal. Our best bet is that RankBrain is somehow helping Google better understand and classify pages based on their content. Perhaps RankBrain can better summarise what a page is about than Google’s previous existing system has done. All we’re sure of though is that RankBrain is Google’s 3rd most important ranking factor. And that’s huge.
How RankBrain affects SEO
We’ve already discussed that RankBrain is helping ambiguous search terms be translated to shorter, more accurate ones to provide more relevant results. Therefore long-tail keyword strategies may take a small hit but if you’ve been following traditional best SEO practices, then for now at least it doesn’t seem like RankBrain calls for any major changes to your current SEO practices. However, there are a number of things you need to get right to be well positioned for RankBrain.
How to optimise content for RankBrain
Niche websites that are laser focused are best positioned for RankBrain. When it comes to writing content for RankBrain, make sure it’s 10x content. Answer your users questions (if you have the best answers, RankBrain will love you), write well-researched, thorough content, answer the questions people aren’t even asking yet (remember 15% of Google queries have never been made before), publish enough content within your niche to become an authority and create evergreen resources. Of course all of these things aren’t specifically geared towards optimising for RankBrain, but these are the foundations of good SEO content.
How to optimise keywords for RankBrain
As RankBrain essentially just does a better job of matching user queries with results, your job is actually going to get easier, not harder. As Google, through RankBrain can now better understand synonyms, homonyms, etc. it’s likely you’re not going to have to work as hard putting all the words from user queries on your page/s. Of course you’ll still need to do keyword research so that you know how best to target a page to your niche. All this means is that as always, it’s getting more and more important to use natural language on your web pages.
Machine learning and even deep learning is going to continue to grow in importance. Whether with RankBrain or another artificial intelligence, Google is always going to rank websites by what the people want. This is and always will be good, relevant content. RankBrain is just another means of getting this content to the top of the search results.
Want to store this information? Get our PDF version