seo-nerd® – digital success
Kurfürstendamm 11, 10719 Berlin, Germany


Artificial intelligence | Stand: Nov. 2022
reading time: 6 minutes

The SEO world has gained one more cool term, enterl and applause please for RankBrain. A new ranking factor is called that which is calculated with artificial intelligence. Sounds exciting? Well, it is! Particularly since RankBrain is already the third most important ranking factor altogether, according to Google.

What does RankBrain do?

RankBrain translates human languages into mathematical entities. It deals with vectors, to be more precise, that are calculable by computers. These vectors represent our understanding of words and terms quite well. When reading the word “Berlin” for instance, most of assume it has to be about the capital of Germany. When discovering terms like politics, chancellor or parliament in the same text, we consider our assumption directly confirmed, even if we only scan the text superficially. We can put the term without effort into context with Madrid, Washington or Rome and would be surprised if Mettmann would come up in this list, as well.

A word thus has more than dimension in regards of its meaning. Vectors can capture the dimensions of a term. And not only that: in the vector space of a term you can also capture, how often this term occurs in documents. Exactly that makes working with the so called term vectors so interesting for Google.

After all, the search engine gathers on one hand huge amounts of written language which it reads with the aid of crawlers and bots on websites worldwide. On the other hand there is also an incredibly large amount of written search requests (to make it simple we’ll just assume that Google Now voice queries are also processed in written form).

The collected language material is translated into term vectors by RankBrain. These can be easily compared for similarities. The more similar a vector of a search request is with the vector of a language pattern on a crawled site, the better the site fits the search requests.

Correspondingly, it can rank higher. Google’s nickname obviously refers to this skill. RankBrain gives matching ranking positions for the search request.
Similarities of vectors are not equating with matches of words and phrases, by the way. The “understanding” of terms captured in vectors is a lot deeper. That way it is also possible for RankBrain to deliver suitable results for queries never entered before. To understand even better how this is achievable it is useful to consider RankBrain…

  • … a machine
  • … as part of the entire search algorithm
  • … isolated as ranking factor, as well


The RankBrain as self-learning machine

RankBrain is constructed in a way that it learns from mistake itself and can do optimisations on its own. Hence, it is a self-learning system. The so called machine learning is used for that, a discipline of AI, which is also the reason why we talk about a self-learning machine.

Since Google only rarely talks about its ranking factors, it is probably only due to the self-learning skill that we have learnt about RankBrain at all. After all, Google’s mother company Alpha invests a lot of time and money into the development of AI (e.g. the lighthouse project of self-driving cars). In this respect, the announcement of RankBrain was surely NOT the beginning of the disclosure of all ranking factors, but rather PR in terms of artificial intelligence.

The machine works autonomously only to a great extent, thought, according to Google. It is monitored carefully and fed with new data regularly, contributing new concepts to the machine testing on its own.

These are first tested offline though, before they go online.

Google has developed the machine over years and has thoroughly tested it ever since, for instance with A/B tests. They found out that user search results that were generated with RankBrain were evaluated more positive and matching than those without the machine. It could even beat its colleagues from Google without an effort. The search engine experts looked at random sites on the net for that and were supposed to rate them on which position the site should rank. Their match quota was tidy 70 percent, RankBrain got even 80 per cent, though.

Is RankBrain behind the ominous Phantom Update early 2015?

The suspicion kind of suggests itself immediately. RankBrain was rolled beginning of 2015, according to Google. Exactly at that time SERP dances could be observed worldwide, i.e. up and downs in the ranking of sites. Since Google, as always, keeps silent and correlations didn’t make a clear picture the changes were called a phantom update. Until mid-2015 the hints accumulated suggesting a core update of the algorithm.

If – and evidence supports this – RankBrain should be behind the update the suspicion from the summer wouldn’t be correct, but not really false either. For the core, the heart of the algorithm was not adjusted or rewritten as assumed then. The search algorithm was apparently supplemented with a new, important ranking factor, though: once again enter and applause for RankBrain, please.

The previous conclusions of the SEO scene were close to the changes of the calculation of ranking positions. It was assumed that Google evaluates the intentions of queries in a completely new way and the changed rankings were according to this.
Exactly this is confirmed with RankBrain now. From the user’s point of view the machine refines the search results.

The RankBrain as part of the general search algorithm

RankBrain works so well apparently according to Google, that only two factors are more important in the search algorithm. What factors these are Google still doesn’t say. The SEO scene mainly agrees though, that both top positions can only be site contents and probably backlinks. Backlinks might have a lot of competition from user signals, though. Factors like bounce-rate, visiting duration or click-through rate (a measurement for the number of sites on which are surfed in the clicked search result) are weighted more and more in the last years.

User signals are helpful indicators for how well a search result fits a search request. Backlinks are one of them as well, of course. If a site is often linked, it cannot be so bad. The disadvantage of backlinks is that they are highly manipulative. With the Penguin update Google tries to combat this problem, but backlinks still are problematic from Google’s viewpoint. By introducing RankBrain Google breaks away once more from its previous dependence on backlinks; and this simply by adding another factor into the algorithm.

RankBrain is the mediating entity between search requests and site contents, so to speak. With RankBrain Google improves the semantic search as well, after all. It acts as the benchmark of the Google algorithm since the Hummingbird update in 2013. Google appears as semantic search engine since then. Behind that the not-so-humble claim hides a search engine to be that understands search requests as well as site contents as humans understand them. In simple words, you may imagine the work of the search engine as follows (cf. the infographic below):

The Hummingbird algorithm decides over the ranking of sites as last instance. For that it refers to the data of the crawled sites, considers backlinks and user signals of previous search requests and gets support from RankBrain recently, with which Hummingbird can make better assumptions about what the searching intends with his or her query.

RankBrain as Ranking Factor. Or: What does this all mean for website owners?

Since Google named RankBrain a ranking factor, the question pops up what this means for website owners and for SEO. The answer for that is luckily simple and follows the route Google has chosen years ago:

  • By RankBrain content marketing for websites becomes even more pivotal. Even more frequently than before content will rank that is relevant for the search request.
  • To serve the semantic search by Google optimally, websites need especially a lot of good texts. These should be optimised for the focal keywords as well as the synonymous terms.
  • Website owners and SEOs have to question and research very well with which requests could be searched for the specific offer. Even if a site is optimised on one keyword (or a long-tail keyword), topic relevant terms should be included on the site and dealt with in the text.

Comments are closed.

    Ihr Warenkorb ist leerZurück zum Shop