What is RankBrain

RankBrain

RankBrain: Brief explanation

RankBrain is part of Google's algorithm that is used to process search queries and determine the SERPs. However, RankBrain is not just any bit of an algorithm, but a subsystem that contains the first approaches of artificial intelligence and can therefore learn from the search queries. This helps the search engine giant to better answer unknown or complex queries.

Detailed explanation

RankBrain was introduced by Google in 2015 and quickly became an important factor in the selection of search results. In view of the fact that there are around 200 "signals" for Google that determine the ranking of a page, this is quite astonishing - but only at first glance.

Because RankBrain can show its strengths, especially when answering search queries, which have not yet been asked. Of course, one could think that with several million inquiries per day, every possible variant has to end up in the Google search field at some point. That would be a fallacy, however, because according to Google, around fifteen percent of all incoming search queries have never been made in this form. This in turn can lead to problems with "normal" search algorithms. How is the request to be classified? Does it contain slang or ambiguous terms? How are these to be assessed? These algorithms can hardly deal with these questions and, in the worst case, deliver completely unsuitable search results.

RankBrain, on the other hand, can draw conclusions and thus also classify unknown words and phrases correctly with a high probability. It is also capable of learning, it saves new terms and can include them in subsequent search queries - among other things, to recognize the meaning of other, unknown terms. This machine learning makes RankBrain something very special among Google's algorithms, because it is not a static system that always evaluates data according to the same scheme. It changes and is expanded with every unknown request. That makes it very powerful - and quite problematic from an SEO point of view, since there is no longer a fixed set of rules that is used to evaluate pages.

How does RankBrain work in detail?

Similar to other Google updates, the search engine giant is of course largely covered with RankBrain when it comes to functionality. In terms of its circumstances, however, Google at RankBrain is unusually open-hearted: In an interview with Bloomberg, Senior Research Scientist Greg Corrado shows how RankBrain works.

Apparently the algorithm is able to break down the entries in the search field into so-called "word vectors" using a mathematical system. In this way, words and the relationships between them become measurable and regularities can be derived. These in turn serve as the basis for the evaluation if RankBrain encounters words or word combinations that it does not know. If a term is not found in the database, the algorithm can "guess" the meaning based on the relationships and regularities established so far. For this purpose, Google at RankBrain relies on a conversation model for processing the requests, in which the entire request is viewed as a unit, instead of treating words individually. The model is supported by a so-called “sequence to sequence framework”, which uses previous sequences (inquiries) to help determine the next sequence.

At RankBrain, machine learning and artificial intelligence are combined into a system that becomes better and more reliable over time. Of course, this required considerable efforts in advance, because Google first had to create a broad database for RankBrain. Without them, the system would have no basis on which to process new search queries. Obviously, this basis was already quite good at the start of RankBrain. During an initial test phase, RankBrain's results were compared with those of normal users. The human users were able to correctly match around seventy percent of all search queries. The machine, on the other hand, came in at eighty percent - and has since gotten even better thanks to the incoming searches. It is therefore hardly surprising that Google RankBrain attaches such great importance to the compilation of the SERPs.

How “intelligent” is RankBrain actually?

Although Google itself repeatedly speaks of artificial intelligence in connection with RankBrain, the term is often equated with machine learning, which is not correct in terms of content. The fact that machine learning is used at RankBrain is undisputed - but how much artificial intelligence is there in the system?

It is difficult to find an answer to this question, because there is neither a generally applicable definition of intelligence, nor can the term artificial intelligence be precisely narrowed down. Most of the time, intelligence is understood to mean the ability to solve complex problems or to gain more in-depth knowledge through logical thinking.

RankBrain is of course still very far from that, because it evaluates the incoming search queries, but cannot understand or interpret them.

Nevertheless, at RankBrain you can - at least in rudiments - speak of artificial intelligence, because the system is able to evaluate inputs based on a certain classification and classify them on the basis of previous "experiences". It will certainly be many years before it becomes a real artificial intelligence, but it is a first step - and a very important one for Google. The search engine giant has not concentrated on its core competencies for a long time, but is increasingly focusing on the development of intelligent and self-learning systems.

RankBrain is therefore not only a way for Google to improve the performance of the search engine, but also a research project that the company can build on in the development of further systems.

RankBrain and its importance for search engine optimization

As already mentioned, RankBrain is a decisive factor in the compilation of search results - according to Google, it is even the third most important. Of course, SEO experts are also listening.

The question, however, is whether you can influence a system that is designed for complex and previously unknown search queries in such a way that your own page slides up in the search results. Classic SEO elements such as keywords, text structure and backlinks ultimately come to nothing here. Instead, RankBrain focuses on semantic relationships and content. With RankBrain you will hardly achieve anything with simple “trickery”. Rather, the system gives an outlook on what will be the focus of more and more websites in the future: structured content with high thematic relevance. Because only if the “word vectors” of the page match those of the request, the page can make it to the top in the SERPs.

How exactly this can be achieved is still open, but the growing importance of content has already become apparent in previous Google updates, and this trend will certainly be reinforced by the successful use of RankBrain.

Conclusion

With RankBrain, Google has for the first time integrated a system into the search algorithm that can learn independently and thus improve with each use. The great success of this system apparently came as a surprise to the company itself, but it should ensure that Google further intensifies its efforts in the field of artificial intelligence.

RankBrain became the third most important factor in ranking shortly after its introduction, but due to the high complexity of the system it is difficult to use it effectively for search engine optimization. However, a clear semantic structure and high thematic relevance should in all probability help to achieve a good ranking via RankBrain. Since content is becoming more and more important for Google when evaluating pages, this is a promising approach in any case.

Join over 53,000 customers worldwide and use Textbroker for your content marketing.

register now for free