Initially, Google classified documents according to the links’ strength and how it interpreted those links, things which are a little complicated but still "understandable" for humans like us 😊. Today, it's a bit of a catastrophe, and especially since RankBrain: artificial intelligence is everywhere, and it’s a question of understanding the broad outlines in order to keep ranking high on Google.


This article is very much inspired by François Gouble, founder of Oncrawl, at the SEO Camp 2017.

(Quick) history of patents filed by Google

Come on, let's go through a very brief history, just to understand how “we got there” :):

  • 1997: Page Rank
  • 1998: Semantic improvement through anchors
  • 2007: ranking of Internet users' searches according to intentions: navigational, transactional, informational keywords...
  • 2009: navigation queries’ improvement & queries related to brands’ rewriting. "Boorish" link with Google shopping
  • 2010 :
    • Google moves on to metadata: it associates Arnold Schwarzenegger with Terminator, Senator, etc. -> we started to go beyond the keyword
    • Co-occurrences: no need to have the exact keyword anymore!
    • Rich snippets: you know, when it looks good on Google results, with rating stars and all that.
  • 2011: the machine learning integration’ starts with Panda
  • 2012: Knowledge Graph & Hummingbird: Google answers your questions
  • 2016 :
    • Zero position: we'll talk about it again!
    • RankBrain: Machine Learning process aimed at refining its results. One of the 3 most important ranking factors according to Google

Focus on Rank Brain?

A few quick words about Machine Learning:

  • The user does a search
  • Google lists out related results
  • The user clicks on a few links, not all links. He/she spends time on some pages, not others: RankBrain analyzes all this, and learns from it. So, the next time a user does the same search, Google will have modified results based on former users’ behaviors on that same query -> ouch, we are far from pure netlinking, and on-page there, aren’t we? 😊

A quick example?

Come on! 😊 On the word "apple" or "Apple": it's the same thing, right? :) Well, no. Actually, Google will understand very easily, "thanks to experience", that in one case users wanted to know about granny apples, and the other about Apple products. See below another more detailed example focused on user feedback, because that’s where the issue is today, especially on click-through-rates.

Focus on Click-Through-Rate (CTR)

We just saw above that Google learns the behavior of Internet users to build its ranking. So, do we need to look at our own website CTRs?

Ouch, not cool, how do I do that?

In your favorite tool, the Google Search Console :).

We go into the research analysis, and we look at the average CTRs.


3 tips to optimize your CTR

Let’s google "men's shoes", and try to understand.


  1. Optimize your titles !
    • Stop the exact match between the page’s main keyword and the title. Let’s send some nice marketing slogan that makes users want to click
    • On men's shoes: the second result is quite cool, isn't it? It tells us why we have to click on it: “Men’s Leather Shoes” for Finsbury Shoes, "Low Price Offer..." for Ajio, “Discount up to 50...” for Myntra. “Men's shoes” from Deichmann is less attractive.
  2. The rich snippets: sometimes there are stars (reviews) under the page title and the website’s link, it’s reassuring. 😊
  3. Does the page contain “entities”: if a page talks about Lady Gaga, Google looks to see if there are any Lady Gaga co-occurrences on the page. Okay true I agree, that's a topic on its own, right... 😊 We'll talk about it again!

Focus on usage metrics

The famous bounce rate, the time spent on site, the loading time...


Yes, before that, it was information only used by us about our users.

Now Google seriously takes it into account, and in real-time with RankBrain: so, if you suddenly notice a very significant bounce rate drop, or a big increase in pages’ speed, it's great news! Google takes it into account very quickly and off you go!
François Goube gave a very interesting example of a client positioned on competitive SEO requests, who advertised a lot. They had a technical issue which removed the ads for a week, and dramatically improved the bounce rate: what a joy! They climbed up spots immediately 😊. It was not intentional, but this kind of information is very important to know, because Google seeks to satisfy users, and it does not satisfy users who leave the search result page immediately. And since RankBrain is running in real-time, this type of improvement can be taken into account very quickly!

Come on, we understand that Google Analytics must be one of your best friends. You have to understand users for Google to like you. Did you say SXO? 😊


Now you know how Google's RankBrain works, and tips on how to master the Google Search Console.

   Article written by Louis Chevant

Further reading

The complete guide to Internal Meshing

The step-by-step method to build your semantic cocoons, your mesh and the optimal tree structure of your website.