How much do you understand about Google’s RankBrain and how it impacts search rankings in 2018?
If you thought that the most important search engine in the world had already stalled with the update of each new algorithm, you may be surprised to learn that this is not the case. Today, more than ever, Google is on the cutting edge of innovation in search and information discrimination. Google uses a machine learning technology called RankBrain to help deliver your search results. This is what we know about it:
Google uses an algorithm with self-learning capability. With this statement, the neuroscientist Greg Corrado, promoter of the Google Brain project, surprised us during an interview with the Bloomberg agency on October 15, 2015, where he confirmed that Google relies on artificial intelligence (AI) to interpret queries from users using a system called “RankBrain”, already integrated in the search engine algorithm months before the publication of the interview.
In fact, it has long been known that Google has spent years investing in the research of artificial intelligence. In 2012 the Californian company recruited technology visionary Raymond Kurzweil as director of engineering, a gesture followed by the acquisition of the AI systems startup DeepMind in 2014 for around 500 million euros.
The so called RankBrain uses by Google is an artificial machine learning system to help sort through its search results. Surely you wonder how that works and how it fits Google’s general ranking system. Below we explain what is known. To begin … lets start with it and how is it related to search algorithms?
With RankBrain, Google’s efforts in the field of AI research go a step further and enter the main business of the company.
Now, is this new technology so smart? What effects does the RankBrain algorithm have on day-to-day webmasters and SEO experts?
What is RANKBRAIN?
RankBrain is the name given by Google to an artificial machine learning intelligence system that is used to help process your search results, as reported by Bloomberg and also confirmed by the same Google giant.
It is the new machine learning AI system designed to process search queries on Google in an entirely new way.
In general terms, RankBrain is part of Google’s global algorithm search; a computer program used to sort through the billions of pages it knows and find those that are considered most relevant to particular queries.
To begin, we must be on the same page with basic concepts: machine learning is where a computer is taught to do something, instead of being taught by humans or after detailed programming. For its part, true artificial intelligence, or AI for short, is where a computer can be as intelligent as a human being, at least in the sense of acquiring knowledge, both to be taught and to build on what it knows and do some new connections.
The last algorithm created by Google for this purpose, is called Hummingbird. For years, this one did not have a formal name. But in mid-2013, Google reviewed it and gave it a name, finally Hummingbird.
According to Corrado, RankBrain has been used since early 2015 as a component of the Hummingbird search algorithm, although initially it was limited to unknown queries. Of the nearly 3 billion queries processed by the search engine on a daily basis, around fifteen percent consists, according to Google, in words and combinations still unknown in that form by the algorithm, which include even colloquial terms, of new creation or complex long tail phrases.
In 2016, RankBrain has expanded its field of action to all Google searches, so today this tech is involved in the processing of all queries entered in the search engine.
In particular, we know that RankBrain is part of the general Hummingbird tool because the Bloomberg article makes it clear that RankBrain does not handle all searches. And out of nowhere, RankBrain, has become what Google says, is the third most important factor in their overall algorithm and for the classification of web pages (just behind links and content).
This is because, unlike previous versions of the Google algorithm, it’s very effective to analyze a search query and return the most relevant results even when it does not contain the keywords or phrases used in the search term. Since it has been deployed, RankBrain has become the most important signal that contributes to the result of a search query.
The main task of RankBrain is basically the interpretation of the words and phrases of each query in order to deduce the intention of the user. In the aforementioned article published by Bloomberg we talk about RankBrain as an artificial intelligence system capable of learning by itself.
But what does Google mean by artificial intelligence and how RankBrain works exactly?
AI research is reflected in this definition of computer science Elaine Rich: “Artificial intelligence is about studying how to make computers do things that, for now, people know how to do better.”
True AI only exists in science fiction novels, of course. In practice, AI is used to refer to computer systems designed to learn and establish connections.
This leads us to clarify the distinction between the two types of artificial intelligence, the strong and the weak:
Strong Artificial Intelligence: to correspond with this definition, a machine should possess intellectual qualities similar to those of humans to be considered intelligent. In addition to the ability to draw conclusions and solve problems, here are qualities such as consciousness, self-knowledge, self-perception, sensitivity and wisdom. The goal is to create intelligence.
Weak Artificial Intelligence: according to this concept, it is enough to equip a machine with qualities associated with people with intelligent behavior. The objective would be, in this case, to imitate with mathematical rules AND intelligent human behavior with qualities such as logical thinking, decision making, the ability to plan and learn and communication.
When Google refers to RankBrain as an AI system capable of learning it does so based on the weak definition of artificial intelligence, since it is a technology that seeks automatic solutions to problems that until now were processed by people. Like most systems of this type, RankBrain also resorts to machine learning techniques.
By machine learning or learning machine it is understood the artificial generation of knowledge from experience. The machine learning systems analyze large volumes of data, identify patterns, trends and transversal connections with the help of mathematical algorithms and based on them they make their own predictions.
How does RankBrain work?
Thanks to RankBrain, the relevance of the content and the intention of the user are once again the focus of optimization for search engines.
RankBrain is used by Google to interpret user queries and select in the Google index, a database of about one billion gigabytes, exactly those web pages that are closest to the user’s intention. For this, this AI system goes beyond the mere collation of key terms.
It is primarily used as a way to interpret the searches that people present to find pages that may not have the exact words that were searched. Google has found pages beyond the exact terms that someone searches for a very long time. From related concepts to synonyms, even if the context is different.
The Knowledge Graph, launched in 2012, was a way in which Google grew even more intelligently about the connections between words. And most importantly, it learned to look for things that were not chains.
The Knowledge Graph is a database of facts about things in the world and the relationships between them. That is why you can do a search for people, things or related concepts, even if you do not give context if you do not write exactly what you want to find.
Does not it sound wonderful? This is how the Artificial Intelligence (AI) or RANKBRAIN that we talked about works.
With the update of Hummingbird in August 2013, Google implemented the so-called semantic search. Whereas, before this update, the terms and combinations of words that were used to search were evaluated statically without a meaning relationship, with the aforementioned update of the algorithm the meaning of the search comes to occupy a central place.
With RankBrain, Google completes the semantic search with a system capable of learning by itself and of using acquired knowledge to respond to new and unknown searches.
Before Bloomberg, Google exemplifies the performance of RankBrain with this search:
“What’s the label of a consumer at the highest level of a food chain?”
Instead of analyzing each word in isolation, RankBrain records the semantics of the query with a global sense, thus determining the user’s intention. This is how, despite being a long tail search, the user quickly gets an answer to his/her question.
To fulfill its task, RankBrain uses the experience it has accumulated with previous queries, creates connections and relies on them to predict what each user is looking for and choose the best way to respond to their query.
To do this, it tries to resolve any ambiguity and deduce the meaning of terms that are unknown until now as neologisms, for example. Obviously, Google does not reveal how it faces this challenge, although SEO experts suspect that RankBrain translates queries with the help of vector word representations in a way that allows computers to deduce semantic relationships.
Already in 2013, Google launches Word2Vec, an open source machine learning software with which to transform, measure and compare the semantic relationships between words in a mathematical representation. In order to carry out this analysis it is necessary to have a basic linguistic corpus.
To “learn” the semantic relationships between words, Word2Vec generates a multidimensional (n-dimensional) vector space in which each word of the underlying corpus is represented by a vector in as many dimensions as the n value defines. The more dimensions that are selected, the more relationships with other words the program can recognize.
In a next step, this vector space is introduced into a neural network that allows, with the help of a learning algorithm, to adjust the vector space in such a way that the words used in the same context generate a similar vector. The similarity between vectors is calculated from the so-called cosine distance as a value between -1 and +1.
In short, if a linguistic corpus is provided to Word2Vec as an input, the program delivers the corresponding vectors as output, which allow evaluating the proximity or semantic distance of the words contained in the corpus.
If the program is faced with a new input, the learning algorithm enables it to readjust the vector space and create new semantic connections or reject previous conclusions. This is how the neural network “learns”.
Officially, Google does not reveal any relationship between the way Word2Vec works and the RankBrain algorithm, but unofficially it can be assumed that the AI system is based on similar mathematical operations.
How does RankBrain help refine queries?
The methods that Google already uses to refine queries tend to flow back to some human being somewhere that does work, either having created derivation lists or synonyms lists or making database connections between things.
Of course, there is some automation involved. But to large extent, it depends on human work, not everything is AI. The problem is that Google processes three billion searches daily. In 2007, Google said that between 20 and 25 percent of those queries had never been seen before. IN 2013, it reduced that number to 15 percent.
But 15 percent of the 3 billion is still a large number of queries never entered by any human seeker – 450 million per day. Among those that can be complex, the queries of several words, surely.
RankBrain is designed to help you better interpret those queries and translate them effectively, behind the scenes in a way, to find the best pages for the search engine.
As Google told us, you can see patterns between complex seemingly unconnected searches to understand how they are really similar to each other. This learning, in turn, allows you to better understand complex future searches and if they are related to particular topics.
The most important thing is that you can associate these search groups with the results that you think searchers will like the most.
Google has not provided specific examples of search groups or any details on how RankBrain guesses which the best pages are. But the latter is probably because if you can translate an ambiguous search into something more specific, then you can get better answers.
Therefore, if you have ever been disappointed with the results of a search, you may want to think about it better. It is not as simple as it seems! Behind all this magic, there is a lot of work from thousands of people, from millions of websites, from the development of an algorithm (or several of them) that despite being an obstacle for SEO experts is a perfect lock for those who wish to advance with traps.
Today, more than ever, Google takes care of its users.
RankBrain and SEO
Artificial intelligence (AI) has been the most trending topic in the SEO industry, and it is also greatly misunderstood by the site owners or webmasters.
One of the most important things to understand about AI is that, it isn’t based on a static formula. Rather it is always on changing and is a dynamic system designed to identify, sort, and present the data – that is most likely to match exactly the needs of users at the time of searching, based on number of variables that go – far beyond just a simple keyword phrase.
Artificial Intelligence is trained by using the following known data, such as:
- User behavior
And by analyzing the above data using the UX, big data, and machine learning in order to develop new ranking factors that is capable of producing the best results most likely to match user needs.
The present SEO is all about the following three major factors:
- Technical SEO
- How to write great engaging contents
- How to develop authority links
In technical SEO, you have to take care of keyword density, UX and make it easy for search engines how easily they could navigate your site and understand its contents. Components like loading time, bounce rate and page responsiveness play a very strong role.
They say content is king, here I would little bit modify this term and say it like that “users engagement is king”. No matter how good content you write, and how long it could be with number of high volume or long tail phrases, at the end it won’t work and produce you great results.
Now-a-days people prefer and are looking for great engaging contents. Therefore what you want to add to your site, think twice about the users experience, later on think of search engines. If not so, then I think you can’t outrank your competitors in SERPS.
Also to mention here that making high number of inbound links using different techniques will adversely affect your site’s authority, and in some cases chances are that Google penalize your entire site out of ranking. That sounds risky!
AI or Artificial Intelligence is a game changing for Google and it totally revolutionize the SEO. It is not based on a static formula; rather it is something always on changing and dynamic. It notices UX, big data, and after all using machine learning to produce exact matching results in accordance to user needs, at the same time learning and improving the results day to day.
So, it means instead of doing keyword stuffing into any content or title, try to focus more on writing something natural, something that looks organic such as you tell a story to your friend or start a debate with him about something. Doing so, you just don’t focus on keywords but you say what it comes to your mind. And just suppose that there isn’t exist any organic search term in Google.
We all know that Google ranks a page according to its content, but in some circumstances, it also ranks a webpage based on some info that even isn’t available on the page. That’s why you have to include related search terms and ideas as naturally as you can … as it will add more value to users and will be positively granted by Google.
The other important thing is to know the search intent of a user:
If you want to beat your competitor, you need to know about the search intent of the visitors. Sounds strange! What is that? You just don’t need to follow what the visitors are looking for, but you have to guess and find out why are they looking for it?
In other words, you have to study the visitors’ brain and heart, what is their main problem and what answer could satisfy them well.
What to think Next?
It isn’t the end … you need to predict what could be the next question of a user after his/her first search?
If you search for “fbi agent from wolf of wall street,” you will see tons of information in Google SERPS such as on top organic search results you will see Wikipedia followed by Imdb. You can easily say that both of the websites have more domain authority due to their high quality authoritative links and contents.
But let me ask you, do you really think so? – I mean the reason behind their ranking?
Honestly I don’t think so.
Here I’ve to prove it – just think about different angles you could navigate from a single source. For this let’s take an example of IMDB. From The Wolf of Wall Street (2013) movie page on it, that already has sufficient amount of content about the movie, you can also find useful links to other important info about:
- The cast and crews, directors, products and writers etc … behind the movie.
- Not only this but you can see list of all of other movies and dramas as well.
- Different reviews about the movies as well as third-party sources.
- Top movies or similar movies and the best tv programs.
Just imagine how it all interconnected to each other, so almost any question or query that could cross your brain as a result of your first search can be seen on IMDB site without leaving it. Sometimes I find myself deep inside it that even I pass 20 to 30 min surfing the pages one after another – and this is what I call “engaging content is king”.
They really done a nice job of predicting about visitors next search intent.
IMDB does content development on a huge scale, as they are targeting a highly competitive market, so better for them to do so. But did you ever think what if you apply their SEO and marketing strategy to your business?
Just for your convenience, think about it like this – the AI main target is to deliver information and precise content to match the users’ search, so tell us which of the following make more sense:
- Send the users to a special site just because it has high authority or
- Send them to a site that not only contains what they are looking for, but also contains the type of other search queries which the users will tend to look for after entering the site via their first search.
Think of a superstore when you go there to buy a shampoo but later you feel why not to buy some bath soaps, tooth pastes, brushes etc … if you can find all the best stuff in a single store why to go to other store?
Even more surprising that making public that the results of Google’s AI research influence web search is the way they do it: Google not only allows RankBrain to interpret all queries since 2016, but according to Greg Corrado, a senior research scientist at Google who is involved with RankBrain:
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.
RankBrain has become the third most important ranking factor in Google’s search algorithm.
For web administrators and SEO experts this requires a change of perspective on the keyword strategy because Google, as a semantic search engine, is in a position to use acquired knowledge, in the form of concepts and connections, to deduce the meaning of any query.
This means that if a web page positions well for a given query it does not depend so much on whether it contains the specific search term as on the relevance of the content of the web page with respect to the concept that RankBrain links to that key term: the priority is not concentrated both in the keyword itself and in the relevance of the content of the web page.
This conclusion is also reached by the software company Searchmetrics, which is present in international SEO with a search and content marketing platform and since 2012 publishes a series of reputed studies on the central factors of Google algorithm ranking.
The classic general and universal list of SEO factors, as we know it up to now, is obsolete if we pay attention to the central conclusion of the last study, which dates from 2016. Searchmetrics examines the latest changes in Google to reach the conclusion that the ranking factors, understood in a generalized way, no longer correspond to the current state of the evolution of the search engine.
Web administrators should focus more on the relevance of the content and place the user at the center of the strategy. However, the requirements of users regarding web pages vary significantly depending on the sector. This has led the company to announce that in the future it will be dedicated to developing differentiated studies.
In order to study the relationship between the ranking of a website and its relevance to the term used in the query, in its latest study Searchmetrics also uses word embedding software that translates semantic relationships into vectors.
Based on a set of 10,000 keywords, the company extracted the relevance of content for each key term of the first twenty results by hiding the keyword. For this, the analysts eliminated the search term analyzed from the texts of the classified web pages and determined for the rest of the content a note of relevance from 0 to 100, which was put in relation to the position in the list of results.
The study highlighted that those pages positioned in the first positions were distinguished by a relevance of much higher content than those worst classified. The highest relevance extracted by Searchmetrics was for web pages between the third and the sixth position.
Do not forget that the first two positions are often occupied by corporate pages that, according to Searchmetrics, benefit from an additional branding factor.
That said: how do webmasters react to RankBrain and the other changes? Searchmetrics mentions holistic editorial design as a central factor for success, that is, a creation of content that, instead of being guided by keywords, does so by themes, always taking into account the user as a priority.
The goal is to respond to queries that are made in Google with relevant content. To do this, web administrators should obtain the search intent for all the keywords with which they want to position. Only in this way can the key terms be structured and grouped into thematic fields that should serve as a cornerstone in the preparation of the editorial plan and in the creation of qualitative and content-rich texts.