What is RankBrain?
Not a long time ago Google released their new secret weapon – RankBrain.
As usual, Google is not telling us much. Everything we know about RankBrain is based on some sort of reverse engineering. We know it affects the SERPS. We know it’s responsible for about 15 – 20 % of the results in the SERPS.
To explain RankBrain let’s start from the beginning.
RankBrain is part of the Hummingbird algorithm and to better understand Hummingbird think about a system that takes into consideration the meaning of the whole sentence, rather than particular words.
Example: 10 years ago the query ‘water pill’ would have probably sent AltaVista or Yahoo to look for a pill with water in it. Nowadays Google understands that this is a diuretic that causes a person to lose water from his body.
When it comes to RankBrain add to the mix a project called Word2Vec. Now I am not a googler and my understanding of this project is that the tool takes the text as an input and produces the word vectors as an output. To better understand the Word2Vec model think of it as a system that incorporates two parts:
One called ‘continuous skip-gram‘ that tries to predict neighbouring words to a given word.
The other one called ‘continuous bag of words (CBOW)’ that tries to predict a word from neighbouring words.
By using those two methods/algorithms Rank Brain can understand or shall I say try to guess the meaning of a word it isn’t familiar with.
So now that we know how it works, or at least how we think it works let’s discuss the name. The first part ‘Rank’ implies that it has something to do with ranking. On top of that, Google said not a long time ago that it’s the third most important signal contributing to the result of a search query. So in a way, it does affect SEO. Saying that it does not affect rankings.
What is RankBrain and how is it Affecting your SEO?
If you are already ranking for term ‘used cars’ (because you sell second-hand cars) and you were number one before RankBrain was introduced, you will still be number one for ‘used cars’ today.
On the other hand, if before the new algorithm you were also ranking for used steering wheels because there is a topical connection with used cars now RankBrain may re-arrange the SERPs and only display companies that are in fact selling steering wheels and not used cars.
So in a way, your rankings haven’t changed but you are no longer present in the SERPs. If you check your favourite software like SEMrush or STATS your website will still be shown as number one for a particular search term but you will no longer appear in the SERPs for terms that are not relevant to your business.
It filters you out of the inappropriate and keeps you current in the appropriate (SERPS) – Zara Altair
A long time ago the way Google worked was that they will crawl your website, gather all the information and when the query comes they will display information based on the keywords in your copy. RankBrain is a shift towards understanding the query first before connecting it to the keywords in your copy. I hope that makes sense.
The second part of the name is Brain and many people are creating link bait articles connecting AI with RankBrain. Machine learning is part of the AI but it is not AI.
What is machine learning?
Machine learning uses statistics to develop self-learning algorithms. In that matter, Google is hoping that in the future RankBrain will be able to run the SERPs without human intervention. Whether this is gonna happen and when nobody can really say. One thing is for sure that the future is algorithms that cannot be gamed.
It is not an “artificial intelligence” as we don’t have that yet. It is a machine learning algorithm. Such algorithms may become a component of any future AI, but they are not yet AI. – Mark Traphagen (Senior Director of Marketing at Stone Temple Consulting)
Why is so difficult to game RankBrain?
To give a concise answer to this question first we have to understand how the new algorithm works.
- Historical search data is processed by RankBrain, which then establishes how to make predictions based on that data – so they take the data (the whole process is done offline), and if they think it’s good enough they will feed it into RankBrain. In a way, if your website is giving ‘crap’ information it would not be fed into the system and you will lose in the future.
- Those predictions are tested and only then they go live.
- This cycle is repeated each time there is a new search in the system.
So the above in simple words. The whole process is done offline, Google decides what the input would be and only after the forecast is approved it goes live so gaming is almost impossible.
As this is a webinar of the relations between Google RankBrain and SEO I will give you some tips to take on board and implement in your daily activities.
RankBrain: 4 SEO Tips You’ll Need to Survive
When it comes to Google RankBrain my first tip would be would be to go and read again the Google Semantic Search book by David Amerland. By reading this book and implementing some of the tips given in the book you will not just improve your website authority on the web, but you will also be doing everything the right way in order to satisfy RankBrain.
My second tip would be to create laser-focused content on topics in your niche. Creating thoughtful and quality content is still just as critical to getting found online as it has always been.
Use keyword research to:
- First, identify your audience – if you do not know your audience than clearly, you don’t know who you’re talking to. The only way to identify who your audience is is to perform a keyword research. Check what terms are popular, have big search volume, in your niche. Then check what keywords people are using to land on your website and try to connect them both.
- Secondly, try and find what your audience is talking about and what your audience is looking for online. In simple words the questions they are asking. RankBrain is all about asking and answering questions. Especially questions that have never been asked before. Try and answer those questions.
- Thirdly, create content that connects with your audience. If your audience likes sports cars and they are also talking about Formula One, buying tips or the British TV program ‘Top Gear’ include all these in your copy. Then the algorithm will think that you are an authority, I guess and hope that you are an authority already, and your information will be fed into the system.
Now more than ever, you’ve got to be focused on user intent, related intentions the user might have, and identifying gaps in your content, and how your quality compares to your competitors for the same query – Eric Enge ( Why Machine Learning Is Revolutionizing Search)
My third tip: As the main direction in the future for Google is voice recognition create content that sounds natural. Write copy that is descriptive and sounds like the reader is in front of you.
My fourth tip would be: do not forget links. Links provide external trust and applicability to your content. This is another way to show RankBrain that your website is an authority in your niche.
Many people say and try to convince us that Rankbrain is the future and would potentially replace all algorithms. This is simply not true. RankBrain would always be a part of a bigger algorithm whatever the name of this algorithm is in the future. RankBrain will always need links to determine relevancy. Links from other authoritative websites in your niche are basically telling the algorithm that you are an authority as well.
Krystyn Dale: Schema and RankBrain correlation.
Omi Sido: Schema is the best way to show Google what your website is about. Everything started with Google trying to create this database of knowledge (Open Graph). With all the knowledge scraped from the web, Google already knows how the words are connected, how web sites are connected and how different niches are connected to each other. RankBrain is basically refining the search, making sure that all the knowledge Google had before and all the ‘new knowledge’ coming to Google is connected and understood. Marking up the content of your website with structured data is the best way to tell Google if your website is relevant.
Other opinions expressed during the conversation.
Artificial Intelligence has been in development for a long time but Machine Learning in my view is actually part of it. Artificial or natural intelligence comes once the “learning” has happened and the information is understood. Actually, the results or decisions that are taken based on the learning show the intelligence!
So the AI research has been going on for a long time and only now we have successfully implemented machine learning which would take us closer to full AI.
In a way, it’s like saying PageRank didn’t have an effect on the rankings during the 00’s. The reality is that it had a huge bearing on the SERPs results during that time and yet quite obviously both these algorithms have the word rank hidden in plain sight within their respective brandings.
I’m quite sure the marketing experts at Google wouldn’t have gone as far as producing a misnomer to throw us off the track ‘:D) In fact, a lot of thought would have gone into naming the RankBrain algorithm to reflect its true meaning just like they did so aptly with PageRank.
Omi Sido, I really could listen to you for hours. You have a gift and a wonderful style for explaining complex concepts. It’s especially valuable that you’re highlighting the connections between how the semantic web and Hummingbird work and RankBrain’s part in that. And how they are working to “see” connections and relevancy between words, entities, search intent- all that juicy stuff.
(Love the pub example- that’s such a great analogy for listening, being able to hear)
And of course, I love your repeated endorsement of our favourite books. ;o)
Zara Altair, thanks for having Omi and for all your additions in the learning here.