RankBrain, the Google AI revolutionizing SEO

Widely tipped as one of the must-watch trends of 2019 in the digital sphere, artificial intelligence is riding high. The world of SEO is no exception to these profound shifts. This is clear, in particular, from the growing importance of the RankBrain algorithm, an intelligent system grafted onto Google that is designed to better decode users’ queries. Its goal? To help the search engine deliver even more relevant answers to users by getting as close as possible to their search intent.

Here’s a look back at the algorithm’s history and how it works, its impact on SEO in general, and the best practices to follow to tame RankBrain.

What is RankBrain?

Introduced back in 2015 by Greg Corrado, a Senior Research Scientist at Google, RankBrain became a foundational pillar of SEO in just a few years. This algorithm is also closely tied to Hummingbird. We can say with certainty that the algorithm has now become one of the most important ranking factors. For many years, the Mountain View firm has invested heavily in research related to machine learning and artificial intelligence. RankBrain is one of its first products brought to market and tested on a large scale.

An artificial brain put to work for your searches

Breaking down the name RankBrain gives us a glimpse of what this famous algorithm is all about. On the menu: a clever blend of automated learning and artificial intelligence that relies on a neural network. RankBrain’s main challenge is to improve the understanding of how users interact with the search engine and why they choose to click on one search result rather than another. Its mission is to translate the user’s query in order to interpret it better.

Thanks to this AI, RankBrain connects several queries that appear different on the surface but converge toward the same results.

Better understanding users’ queries

Beyond the meaning of keywords, RankBrain focuses more precisely on the true meaning of complete queries, and by extension of sentences or questions as a whole. Semantic analysis then comes fully into its own.

Today, Google has become capable of understanding and interpreting queries that have never been made before, through a highly advanced correlation system. Particularly interesting and powerful on long-tail queries, RankBrain is able to re-situate a word or a concept within a corpus of words. Quite a step forward for users, who can now rely on Google’s performance to understand what they’re looking for even better than a human!

How RankBrain works

As a search engine user, there’s no way to know whether one of our queries is handled by RankBrain or not. What we do know is that the algorithm currently handles 15% of queries. This percentage is obviously bound to increase over the coming years.

Search engines: how do they work?

To better grasp how RankBrain works, it’s important to understand how a search engine works.

Before being able to serve up search results in response to a query, Googlebots have to crawl and index billions of pages. For this reason, optimizing your crawl budget is very important for your site’s SEO! Then, an in-depth analysis of content combined with an evaluation based on the main ranking factors makes it possible to establish a ranking.

Source: OnCrawl

To learn more about this topic, check out our dedicated guide to search engines.

On the user side, we distinguish 3 types of queries: transactional, informational and navigational. While Google isn’t yet able to understand content in the literal sense of the term, it can easily extract the main themes and topics being covered. How is this possible? RankBrain detects named entities and builds matrices that are used, in particular, to define the ranking of results. Each type of query therefore corresponds to a different search result. Likewise, the algorithm evaluates the similarity between two pieces of content.

Examples:

• “ Nike Air Max “ is a query used to reach the product page of the Nike brand’s website (navigational query).
• “ Cheap white women’s Nike sneakers “ will lead to results from e-commerce sites selling discounted Nike sneakers, because there is a genuine purchase intent behind this query (transactional query).
• “ How do I care for my leather Nike sneakers? “ will lead the user to pages offering tips on caring for leather and sneakers in general, wiki-style (informational query).

Google’s AI takes over

Until RankBrain came along, Google engineers were in charge of carrying out algorithm updates. But since then, it’s very much the algorithm that does the work. And the results are there: RankBrain delivers results roughly 10% more accurate than an engineer. Technical teams are increasingly betting on this algorithm, which is continuously improved through machine learning to best satisfy users’ queries.

To refine its understanding of search engine users’ habits and expectations, RankBrain runs numerous tests to figure out which results are the most relevant. Just as a web marketing expert will set up A/B testing campaigns on a site to identify which design, which button, which color, which text delivers the most conversions, RankBrain will serve different results for the same query and analyze what works best, what users like most.

Where RankBrain sets itself apart is in its self-learning abilities. Among other things, they help it re-orient its search results in later searches. The goal is always the same: to offer a personalized user experience that’s in tune with the user’s expectations.

Applied to the field of SEO, the notion of AI built into RankBrain is complemented by machine learning, which gives the algorithm all its power.

Machine Learning and Deep Learning: the future of search

Looking toward 2020, the world of new technologies is increasingly shaped by machine learning and deep learning. Today, SERPs are adjusted automatically through machine learning, which relies on all the data collected so far to establish the most relevant ranking.

Deep learning is currently the most promising area of the machine learning universe. It’s based on neural networks that need a large volume of data to teach themselves, like an autodidact. Neural networks were first introduced in the 1930s, but they’ve only really taken off in the past 3 or 4 years, as computers’ processing power has increased.

This very telling diagram highlights how Google works with RankBrain:

Source: Oncrawl

According to Greg Corrado, the creator of RankBrain, the combination of machine learning and AI could be as revolutionary for our society as the creation of the internet. These new technologies completely transform data processing and open up new horizons for search.

RankBrain is currently the most sophisticated and most accomplished algorithm Google has ever developed. To build a machine learning algorithm, an immense amount of data is needed to produce and train the various analysis models.

How do you optimize your site for RankBrain?

To make RankBrain’s job easier and maximize your chances of showing up in search results when this algorithm is called upon, there are a few best practices to take on board.

Favor synonyms over keyword repetition

Behind RankBrain lies a highly meticulous semantic analysis and a genuine notion of NLP (Natural Language Processing). The days when keyword stuffing still had a chance of working are well and truly over!

To increase your chances of ranking your site on SERPs generated by the intelligent algorithm, we strongly advise diversifying your semantic universe as much as possible, while remaining consistent, of course.

The challenge is to understand exactly which types of queries lead users to land on your site and to optimize your site accordingly. With RankBrain, the SERP served on a transactional query is completely different from the one served on an informational query.

Here’s an example of search results on a transactional query:

And here’s the SERP generated by an informational query:

Generally, priority is given to transactional queries for RankBrain’s SERPs. To maximize your chances of ranking, you can use Rich Snippets to highlight prices and reviews. Using Schema.org markup is strongly recommended.

To show up on an informational query, opt for in-depth articles, enriched with reliable sources and bullet-point lists for better readability. An h1 that contains a question can more easily reach position zero.

For navigational queries, carefully study the search volumes of your strategic keywords to optimize your Title tag accordingly.

RankBrain is able to understand the relationships between the various entities, which clearly broadens the range of possibilities. Google accesses an almost exhaustive dictionary—at any rate one updated in real time and continuously—of every possible synonym for each word. It’s also able to identify the links between different words, concepts, groups of words… This makes it possible to arrive at the construction of named entities:

Source: OnCrawl

Tools like Semji also rely on Artificial Intelligence to help you write content that’s relevant for Google and for users. Want to learn more about our solution? Ask us for a demo!

Analyze the content of competitors ranked in Google’s Top 3

Beyond analyzing user behavior in response to the results served on the SERP, RankBrain takes into account the entire path traveled: time spent on the site, visiting other sites and then returning to the first, CTR, bounce rate, clicked links… Google does everything it can to ensure the user finds what they want on the first try and doesn’t have to continue searching. Google clearly wants to establish itself as the tool capable of delivering THE best answer.

With RankBrain, your semantic score is calculated based on competitors’ performance. It’s therefore worthwhile to study which types of competitor pages show up on a given query in order to build your content strategy.

Thanks in particular to Ahrefs, you can easily identify new keywords and enrich your semantic field.

Source: Ahrefs

Let’s take the example of an e-commerce site selling food supplements for dogs. It logically falls within an e-commerce SERP that features sites specializing in products and food supplements for animals. However, hardly anyone makes transactional queries as such in this niche market.

On the other hand, we find that buyers of this type of product tend to reach a site through an informational query such as “ How can I make my dog’s coat shinier? “ or “ How do I treat my dog’s paw pads? “. This data tells us that it’s better to opt for enriched, expert content that covers the site’s topic. On the contrary, the mistake would be to bet everything on product sheets, which will struggle to rank on the key words and queries of this e-commerce site.

Feed the lexical field and semantic field as much as possible

Google’s knowledge grows a little more every second. With more than 450 million new queries every day, engineers have plenty to intensively train their algorithm with!

Before addressing the “ Why? “ question, let’s first look at the differences between lexical field and semantic field. Often confused yet distinct, these two concepts differ in a few subtle ways.

The lexical field brings together all the words that revolve around the same theme. For example, the lexical field of SEO is made up of, for instance: keyword, netlinking, backlinks, tags, site, traffic…

The semantic field, for its part, encompasses all the definitions of a single word that change depending on its context. For example:

• I’m eating an orange -> fruit
• I’m going on vacation to Orange -> city
• I have an orange t-shirt -> color
• I have an internet subscription with Orange -> brand

For each named entity, Google is able to identify:

• the sentences that contain this entity,
• its usual context,
• the words and expressions that most often accompany it,
• the context of use with a second entity.

Source: Oncrawl

Thanks to this mass of knowledge, RankBrain is therefore able to steer search results according to the context of each query.

So, for a given query, Google tries as much as possible to situate the context of the user who formulates the query in order to offer them the most suitable answer.

For each expression you work on, it’s now crucial to weave the richest and most exhaustive lexical field possible around it. This groundwork also allows the algorithm to expand its knowledge of a field and to consolidate certain links, or create new ones.

Strategically cluster the site’s content and topics

To win over RankBrain, it’s important to build a site and its content while taking into account the expectations of Googlebots, of course, but also and above all those of users. This means precisely mapping the different types of content on your site as well as the categories of pages.

Google takes siloing and the internal linking of sites into account better and better. To make crawling easier, you can silo the technical architecture with a funnel-structured mega menu.

A good way to differentiate your content from your competitors’ is to focus more on the topic to be covered than on the strategic keywords that define it. This simply comes down to letting a human write for a human…

To showcase your content further, building it in topic clusters or silos is therefore highly appreciated.

This process takes time, which is why tools like Semji are now able to deliver, in just a few seconds, the main topic clusters to cover in a piece of content. Ask us for a demo and discover how Semji helps you write higher-performing content in less time!

RankBrain’s biggest challenge is to make all the words and expressions used by humans understandable and, above all, interpretable by robots. Indeed, Google is focusing all its efforts on the use of Word Embedding, which consists of vectorizing words through a neural network.

Source: SEO Quantum

More generally, siloing and the semantic cocoon make it possible to gain SEO performance. So why deprive yourself of it?

In 2019, what will RankBrain’s impact on search be?

To the question “ Does RankBrain shake up the rules of SEO ? “, the answer is: Yes, but to a different degree.

RankBrain rests essentially on 3 key concepts:

• the signals applied to queries: it’s more essential than ever to identify the search intent hidden behind a query and to match it with the most appropriate content. Is the user looking for fresh information or, on the contrary, highly expert, in-depth content on a niche topic?
• the signals tied to your website’s reputation: RankBrain will draw on the reputation and environment you’ve built around your site. This indicator serves to determine the context in which your content will be served to the user.
• the keywords associated with each page: SEO is slowly transforming on this front. The technique of grouping logically related keywords within a single page is now seen as more relevant than the famous “ one keyword per page “. Keep in mind that Google is becoming increasingly intelligent and that it will be more and more difficult to fool it. Otherwise, watch out for Google penalties!

Along with content quality and netlinking, RankBrain now ranks among the 3 most important ranking factors for Google. By its very nature, the RankBrain algorithm learns on its own and is therefore destined to transform continuously. Your SEO strategy must follow the same model, that is, be able to adapt quickly to changes and stay in tune with users’ real expectations. You know what you have left to do!