9th of September 2017 by Marcin Chirowski
How Machine Learning is changing SEO
So, how does machine learning is really changing SEO & impact your rankings?
That’s the big question.
Google updates its search engine algorithms at a lighting speed. An introduction of Machine Learning made SEO rules changed forever. In this article, I hope to sparkle a conversation & break down these revolutionary changes in Google algorithm to help the SEO community stay competitive.
Look, you and many other people in SEO community might be worried about Machine Learning. It’s a hard one to understand at the beginning, but when we break it down into pieces it’s starting to be less frightening.
So, let’s do that and explore that subject in more detail. In this article, I break down the most important parts of Machine Learning. What it is, how it works, how it’s changing SEO & finally what actions we can take to stay competitive.
WHAT’S MACHINE LEARNING
Before we can talk how it’s impacting SEO, we first need to understand what Machine Learning really is.
“Machine learning gives computers the ability to learn without being explicitly programmed.” – Arthur Lee Samuel coined the term “machine learning” in 1959. So, it’s not that new, but it’s clearly picking up in the importance.
- Speech Recognition > Google Voice / Amazon Echo
- Face detection, object recognition, speech recognition, handwriting recognition
- Google Translate is now powered with Machine Learning
- Planes on autopilot / Self-driven Cars
- Facebook: Feed, friend suggestions, Ads rankings
- Netflix movie recommendation / Amazon shopping recommendation
- & of course search engines
GOOGLE PUBLISHED A QUICK INTRODUCTION VIDEO TO WHAT MACHINE LEARNING IS
HOW MACHINE LEARNING WORKS
At the top, you can see a very simplified architecture of a machine-learned search engine (source). The part in the red & green is quite well known to anybody doing SEO and it’s a way that search engines work. New Machine Learning part is highlighted in Yellow.
Let’s walk it through step by step:
- What you first do is to give Machine Learning algorithm the training dataset. Google did that by manually uploading it and feeding it with millions of examples they had accumulated from previous years. Training data consists of queries and documents matching together with relevance degree of each match. Alternatively, training data may be delivered automatically by analysing clickthrough logs.
- Then the algo starts learning from the data. What it is it and want to look for.
- New examples are feeding constantly into Learning Algo
- Algo starts to identify the patterns based on millions of examples
- Algo make a prediction and give you a search result > calculated out of the training dataset
As you can imagine that’s extremely powerful and it’s used already in the number of ways by Google & other brands.
In summary, Machine learning is all about learning from examples & observations. Identify the patterns based on millions of examples and generalizing from there.
HOW IT’S RELATED TO RANKING FACTORS
Let’s have a quick step back and explore: Ranking Factors Evolution, as it will paint a full picture of what’s going on.
Not long time ago in 2015, it was more or less clear what are the ranking factors. Moz published their own version, which was brilliant.
.. But, in early 2015, Google began its slow rollout of machine-learning artificial intelligence system called “RankBrain”.
In this interview: Gary Illyes of Google tells us that Google may use machine learning to aggregate signals together.
As of June 2016, RankBrain is being used for all Google queries.
HOW MACHINE LEARNING IS CHANGING SEO
Now, let’s answer the big question.
Yes, Machine learning is evolving and changing SEO rapidly. But, in my opinion, it’s not like that all the sudden we are going to have several new ranking factors in place.
First of all Machine learning hasn’t been yet developed to the point where it can fundamentally disrupt SEO.
Secondly. Some unknown % of the 200 ranking factors are based on machine-learning. Yes, in future, that % will most likely increase, but I don’t think Google engineers will give away whole power to the machine, as it the search results will be beyond their control.
Finally, Google uses it mostly for “coming up with new signals and signal aggregations”, but it’s yet unknown how exactly they are using it. So they may look at two or more different existing non-machine-learning signals and see if adding machine learning to the aggregation of them can help improve search rankings and quality.
So, in my view, Machine Learning it’s not going to turn the world of SEO upside down anytime soon.
We need to face a new reality, where we are going to see increasing uncertainty over:
- Unknown % of ranking factors impacted
- Unknown why something is ranking on Top
- An Ever-Evolving Search Algorithm – Google it’s not anymore a series of updates, but it transformed into an ever-changing machine learning system. The core ranking systems of search engines get updated with newer systems, they will continue to get exponentially smarter
- Machine learning doesn’t eliminate all your traffic overnight (like Panda or Penguin). It’s a gradual process. Like described perfectly by Larry Kim
.. also our distorted view on Artificial Intelligence is far beyond the reality.
Machines are learning much faster than we think.
WHAT ACTIONS WE CAN TAKE
Finally. What actions can we take? How can we stay competitive and make sure our websites keep ranking high?
1 / Focus on user signals: Bounce Rate %, Page Depth, Avg. Time on Site, CTR %
Behaviour & user signals play increasing more important role as Google continue to evolve algorithm into machine learning mechanism. Now, in my opinion, it’s even more important to put more focus on user signals. As to me, it’s got similar principals as Quality Score in Paid Search.
Larry Kim recently published an article on how Avg. Time on Site is affecting SEO and how you can identify the donkeys. He explained how to find your most vulnerable content — pages that are likely to lose organic traffic because they have low engagement.
2 / Find the donkeys & replace them with unicorns
I’m sure you know that your site has some “donkeys”. It’s hard to admit, but we all have pages that really don’t deserve to be ranking in Google. Now, it’s time to get rid of them & replace them with “unicorns” – pages with a true value to your users. Pages that deliver. Pages that users engage with. Pages full of semantic relevance to your target keywords. Pages that are shareable.
Below you can find one of the ways to find the “donkeys” in your Google Analytics. Just drill down to All Pages report in Behaviour section in GA and compare pages to the site average.
Besides, there are many other ways and I would suggest full Content Audit. To review what you’ve got already and how you can improve it.
3 / KEEP CREATING ONLY GREAT CONTENT & PROMOTING IT WELL
Look, I know that Machine Learning is a new thing and it would be foolish to say that: “It’s going to flip the entire SEO industry upside down.” So, don’t forget about technical fundamentals. As well as creating great content and making sure you are promoting it well, as backlinks remain an important Google ranking factor.
The whole subject of Machine Learning is complex and we’ve just scratched the surface in here, but it should give you a good overview of the most important parts of it. I’m also not claiming that I’m an expert in Machine Learning. So, let me know if you spotted something that should be corrected.
Thoughts? Let’s start the conversation in comments below.
Want more? Dive deeper & learn more about Machine Learning:
- https://moz.com/blog/machine-learning-revolution by Eric Enge
- https://martechtoday.com/how-machine-learning-works-150366 – How Machine Learning Works, As Explained By Google – slides from Google Research Team
- https://www.coursera.org/learn/machine-learning/ – machine learning on Coursera. It’s taught by Andrew Ng of Stanford University
- https://www.youtube.com/watch?v=Rnm83GqgqPE - 45min video from Google Developers about Machine Learning: Google’s Vision – Google I/O 2016
Marcin - SEO geek behind that blog post, cracking the SEO code since 2006.