Google always has something new and amazing up its sleeve, and the most current thing that they have been working on is another step in the evolution of Google Search, from machine learning to its brilliant successor, deep learning.
Everything started back in the day when Google’s scientists were working on the search engine’s synonym recognition. With input data consisting of different words that are interchangeable with one another, Google would use that knowledge to better understand what is being searched for.
Next came website translation, where the system was fed with a wide variety of translated documents, with the goal to “train” Google how one language is mapped to another. The result was Google being able to translate sites into languages that none of the engineers knew. Now, the AI machine learning algorithm that is quickly and drastically affecting the SEO world is Google’s RankBrain.
Deep learning explained
Deep learning is based on the concept of a network of digital “neurons” which are organized into layers. Every layer takes higher level features from the data that it is fed, and then passes them on to the next layer. This way, higher layers are able to understand the concepts that are behind the data.
Let’s take images for example. The first layer is given single pixels and is “trained” to recognize shapes from them. The layers above are then able to combine those shapes in order to “comprehend” what the displayed objects are. If it is given a large number of images that contain human faces and are “taught” the concept of a face, the system will be able to recognize faces in any image.
What is amazing about modern neural networks is that they don’t require “teaching”. The modern concept of unsupervised learning
makes it possible for the machine to recognize concepts behind the input data, without being told what it is looking for.
Google’s RankBrain and its AI
RankBrain’s machine learning algorithms are based on Artificial Narrow Intelligence
(ANI), which is an AI focused on one specific thing. ANI has actually been used for a while now – for email spam filters and Google Translate. There are 5 different approaches scientists take with ANI, so they are grouped into:
Google’s Rank Brain belongs to the Connectionists’ approach. This approach is based on the idea that all knowledge is encrypted in the neural connections in our brain. Which is, as we have previously ascertained, the definition of deep learning.
What does RankBrain do?
Rank Brain takes in Google’s already existing core algorithms and then combines them to display the best possible results. It may decide that the most important aspect of a particular search result is its PageRank
. So, while this may work well for the positioning of one website, for another the determining factor could be the meta title.
What this means is that Google has a very specific combination of algorithms for each search result. This leads to the conclusion that today’s website regression analysis has to be based on specific sets of search results, with a specific context in mind.
Website quality and backlinks
What RankBrain also does is determine what the difference is between “good” and “bad” websites. The same way that it combines the algorithms specifically for every search result, it determines the quality of the website based on a particular factor, such as data structures
, templates, CTR, and the like. In other words, it “understands” what the proper settings for the given environment are. Then it compares the site’s structure to the one of a reputable website in the industry and positions it according to that.
RankBrain’s algorithm also has an effect on your backlinks. It is vital that your backlinks are related to your industry, because RankBrain notices the differences between your website and others that belong to your vertical, and penalizes it if your links seem unnatural.
What can we expect from the future?
The one thing that we can expect from Google’s RankBrain and other AIs is that they will continue improving until they have surpassed the capabilities of the human brain. As we have ascertained, deep learning has deeply affected and will keep on changing the world of SEO. Companies that offer SEO services
will find themselves in big trouble if they don’t stay ahead of the ramifications of AI’s improvements of search engine algorithms. What we know is that:
- Every website needs to look up to the structure of reputable websites in the industry
- If a website doesn’t focus on their particular niche and gets misclassified it will suffer from it
- Every environment that is based on a competitive keyword will have to be examined on its own.
What this means is that there will no longer be any loopholes in the system. But it also means that SEO will become even more technical and everyone will have to adapt.