Guest blog by Romano Groenewoud of SEOgeek
RankBrain is one of the most new and interesting algorithms that Google has recently used to get the best search results. RankBrain is mainly used to optimize search results in Google by converting typed punctuation marks into sentences that the computer and Google can understand. In short, RankBrain is therefore just like the human brain: it sees, decodes and organizes, and then gives a hopefully (sufficient) answer. The response from Google is of course, just like in our brain, dependent on the extent to which the information received is correct decoded and organized. Does the search really make sense?
RankBrain is in his best works: it is able to convert unknown words and sentences into meaningful sentences that Google can understand. But there is a slight difference with the human brain: RankBrain is not the main processor of information. RankBrain serves as 'help', and is really only useful when Google's search results are not optimal, or in fact are not properly 'decoded'. When Google does not understand something, RankBrain comes to the rescue. It then works as an artificial intelligence algorithm and organizes the searches in a different way. Ultimately, but that is science for later, RankBrain wants to develop into being able to recognize sentences that actually say the same:
"What are the best movies on Netflix?"
"What are Netflix's top films?"
"Which films can I watch best on Netflix?"
In fact, these questions ask: what is the best thing to watch on Netflix? All the same question. So far, Google does not recognize this, and you get different results for each question, but RankBrain will be used to see that these questions mean the same thing, so that they can produce the same thing.
But for now, RankBrain provides Google with better search results in various ways, and has recently become Google's third most important ranking factor. The question therefore arises:
How to optimize for RankBrain?
According to the Bloomberg report from 2015 and Google spokespersons, RankBrain guarantees 15 percent of Google's never-before-performed searches every day. RankBrain is also a largely independent algorithm (Deep Learning) that develops itself further, without human input. An example of this is, for example, this Deep Learning algorithm created by Google, which gradually develops into a super player:
There is only one thing you can do to optimize for Google: meaningful and understandable texts to write. But optimizing for RankBrain is a bit more complicated. RankBrain is, from what we know so far, mainly a player in the organization of sentences and is therefore aimed at better understanding of language. The answer to the question is therefore short and clear:
There is no way to optimize for RankBrain.
Guessing about signals
In addition to some factual statements from prominent sources regarding RankBrain and Google, there is uncertainty and unfamiliarity with the precise signals and vectors that RankBrain uses to understand complex language. Vectors are the mathematical entities that are the major meaningful texts, and actually form the basis of RankBrain. The clearer the vectors: the better Google understands what is being searched for, and therefore the better the search results will be. A Google spokesperson does not want to make a direct link to Word2Vec, a prominent model for vectors, but it is a widespread suspicion among techies that Word2Vec is the basic model of RankBrain. The Google spokesperson does want to indicate that it is "related to Word2Vec," and that it can therefore be a layer of RankBrain.
The uncertainty about which signals RankBrain uses is provided by Rand Fishkin (co-founder of Moz.com, see here) somewhat contradicted. He states that RankBrain is concerned with the most important signals that Google uses, in particular:
- Freshness of the content (is it a repeat of last year's news, or is the content really 'new new ' and, does it look fresh?)
- Participation (Commitment, an important signal for Google)
- Depth of content (cut the wood, isn't it nonsense?)
And to a lesser extent:
- Keywords and Keyword Matching (keywords within the text)
- Domain Quality (Google does not want to be overloaded by low quality domains)
These are pretty much the most important signals for Google, and RankBrain blends in with this by helping Google know what the most important signals are and how these signals actually "weigh". RankBrain selects and sets certain signals as a priority, but because, according to Rand Fishkin, RankBrain attaches little value to keywords and other signals, optimizing for RankBrain is therefore not just possible. In addition, RankBrain is one of the hundreds of "signals" that Google uses to create an algorithm, and although we know that RankBrain is widely used, it is not exactly clear how big the influence of the other signals is.
Nevertheless, we can somewhat conclude from Rand Fishkin's words that if you look at the most important signals from Google, partial optimization for RankBrain is possible. But then again, that is not common Optimize for Google?
SEO Will do the trick
It will not make many people happy, but: the uncertainty about signals and the models for vectors that RankBrain uses make optimization impossible. It is impossible to find out which words the best vectors use so that you end up a little higher in the search results. So optimizing for RankBrain is difficult. Some people say that this is the # death of SEO, but rest assured: Google says that the SEO magic is not used up and that good and clear writing is still the best. That is also the most important message: deliver good writing and do not expect RankBrain to translate everything into meaningful sentences, content, depth and participation are the most important.