Google explains the differences of neural matching and rank brain, a reasoning framework called “RankBrain” to help sort through its query items. RankBrain is Google’s name for an AI man-made brainpower framework that is utilized to help process its list items, as was accounted for by Bloomberg and furthermore affirmed to us by Google.
Just like most Google algorithm updates, RankBrain is shrouded in mystery. It has been 4 years since it went live in October 2015, but the subject is still surrounded by a lot of buzz and controversy. But the truth is, RankBrain is one of the most essential parts of Google’s core algorithm and the only machine learning system it uses at the moment.
So in this article, you’ll receive answers to the most frequently asked questions, including how RankBrain actually works, how it changed SEO, and how to optimize for it.
How it all started
Just before we get down to RankBrain itself, let me tell you a quick backstory about what it emerged from. I’m sure you can remember that the 10-years-ago Internet was a big old mess — spammy sites used to rule rankings, website owners used to buy links, and SEO was really far from being called a fair game.
But in 2011, everything changed forever as Google realized that quality and relevant results should come first. So the search engine started a white-hat SEO revolution by penalizing and down-ranking untrustworthy sites with its Panda (2011) and Penguin (2012) updates. Right after more or less quality sites started ranking on the top positions, Google set a course for improving relevance.
Back in the day, Google used to look at separate words within a query to figure out the search intent, which didn’t always work out. That is why it came up with the Hummingbird update (2013), which made a breakthrough in semantic search by taking into consideration a combination of keywords as well as their context. However, search results were still far from being perfectly relevant because the algorithm didn’t know how to process unfamiliar search queries that were constantly appearing. In fact, about 15 percent of the queries Google processes every day are new. So two years later, in October 2015, Google introduced RankBrain, the purpose of which was processing never-before-seen search queries and predicting the best result for them.
#1 What is RankBrain?
RankBrain is Google’s name for a machine-learning system that is used to process unfamiliar and unique queries and relate them to already existing searches, providing users with more relevant search results.
Although the algorithm went live in April 2015, it was first publicly mentioned in an interview with Greg Corrado, a senior research scientist at Google, to Bloomberg in October 2015.
Here is how Greg Corrado described RankBrain at the time:
#2 How does RankBrain work?
RankBrain uses so-called “entities”, which are specific objects that Google knows some facts about, like persons, places, and things. With the help of a mathematical algorithm, it then divides entities into more specific word vectors that lead to certain SERPs. Naturally, similar word vectors lead to similar SERPs.
The best thing about entities is that Google has already collected a lot of information about them and can immediately come up with the most precise search results for your query. However, when RankBrain comes across an unknown query, it searches for the vector that is the most similar to the original query and returns the results for it.
Over time, Google refines the results for a search query that used to be unknown based on user interaction and search patterns. Basically, RankBrain analyzes the results that users finally go for after typing the same search query. If it notices that many users prefer one particular search result over others, RankBrain will consider it more relevant and will most probably rank it higher for other queries like this.
RankBrain also shows great results at understanding negative-oriented queries — keyword phrases containing words like “without” or “not.” Back at the time, Google would simply skip such words.
Here’s how Gary Illyes explained RankBrain’s mechanism at SMX Advanced Conference:
“Basically, it’s a ranking factor. It’s part of machine learning. It’s something that tries to identify patterns and bucket data. It looks at data about past searches and based on what worked well for those searches, it will try to predict what will work best for a certain query. This works best for long-tail queries and queries we’ve never seen.
An example might be “can i beat mario bros without using a walkthrough”. Without RankBrain, we give interesting results that don’t meet my needs. But with RankBrain, we can give results that satisfy my question.”
#3 What queries are affected by RankBrain?
In 2015, when RankBrain was just rolled out, it was used in only 15% of all Google searches. However, in 2016, when RankBrain stated showing surprisingly good results, Google’s confidence in the machine learning system started to grow. But still, RankBrain does not process all the queries, specializing mostly in queries that are unclear to Google. As Steven Levy clearly stated, “RankBrain is probably not involved in every query but in a lot of queries.”
The logic behind RankBrain not being involved into processing all queries is pretty simple — when Google is confident about the meaning of a query, RankBrain is of no use to it. It only enters the game when Google can’t understand what a certain query is about.
#4 What are machine learning and artificial intelligence?
For you to get a better understanding of RankBrain, it’s important that you also have an idea of what machine learning and artificial intelligence are. The thing with these two is that they are closely intertwined and, therefore, quite often misinterpreted.
In a nutshell, Artificial Intelligence is a much broader concept of machines being able to carry out tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
When it comes to Machine Learning, it’s an application of Artificial Intelligence that can learn on its own without being explicitly programmed. This is exactly what RankBrain does — it automatically learns and improves based on its past experience.
#5 What is neural matching?
I won’t lie saying that neural matching is probably the hottest buzzword in the industry at the moment. The topic has become really popular among SEOs after Danny Sullivan announced in September 2018 that Google started using neural matching that was affecting about 30% of all queries at the time.
We haven’t received any clarifications on the matter until this latest Danny Sullivan’s tweet where he described the neural matching concept as:
In a nutshell, neural matching is a system that helps Google better relate words to searches in an attempt to provide the most relevant search results.
#6 How RankBrain differs from neural matching?
Knowing that both RankBrain and neural matching are AI-based systems, they still differ a lot. In addition to the above-mentioned tweet, Danny Sullivan provided a great example of how neural matching actually works:
And here’s how Danny Sullivan explained the core difference between RankBrain and neural matching in two simple sentences:
— RankBrain helps Google better relate pages to concepts
— Neural matching helps Google better relate words to searches.
And there’s nothing special searchers or webmasters need to do. These are part of our core systems designed to naturally increase understanding. ”
Based on these only comments we received from Google so far, the core difference between these two is that they perform different tasks. Neural matching’s main purpose is to connect queries to certain concepts, forming what Danny Sullivan referred to as a “super-synonym system.” Then RankBrain enters the game and returns the most relevant SERPs based on historic user behaviour. Please bear in mind that it’s just an assumption, and there are no official comments, proving us right or wrong.
Although RankBrain and neural matching do different things, they still have some things in common — they both do pretty well at understanding natural language and the meaning behind search queries.
#7 Is RankBrain part of Hummingbird?
Hummingbird is the overall Google search algorithm, which is made up of many different parts responsible for certain tasks. RankBrain also operates under Hummingbird, being responsible for processing unique queries — it doesn’t handle all searches, as only a major algorithm would.
#8 Is RankBrain a ranking signal?
A signal or not a signal — that is the question. Fact is, in the same Bloomberg interview, Greg Corrado did call RankBrain Google’s third most important ranking signal. Here’s what he said:
Based on what you’ve just read, RankBrain is a ranking signal indeed. However, this statement still massively depends on what you consider a ranking signal. In a more traditional meaning, ranking signals are certain website characteristics (like keywords on your page, number of backlinks, page authority, etc.) that search engine algorithms take into consideration while assigning rankings. So if we look at ranking signals from this side, then RankBrain is definitely not a ranking signal — it’s not a website characteristic and there’s no RankBrain score (at least nobody knows about it). That is why in my point of view, RankBrain is more of a keyword processing mechanism than a ranking factor.
But if we look from another perspective, considering a ranking signal as part of the algorithm that participates in the ranking process, then RankBrain can be definitely referred to as a ranking signal.
#9 How RankBrain changed SEO?
Now that RankBrain aims at connecting searchers to the most relevant results possible, search intent has become a priority. That is why RankBrain only gives preference to pages that really meet its requirements — answer searchers’ questions, allow transaction (if needed), or provide a comprehensive piece of information on the topic stated in the query. So basically, today’s efficient content optimization is impossible without understanding search intent and carrying out intent-specific keyword research.
Another thing that has changed the way we do SEO for good is the content optimization focus being shifted from keywords to topics. I guess that it comes as no surprise to the majority of SEOs that the one-keyword-one-page concept is really dead. It means that in the RankBrain era, all you need to strive for is comprehensiveness — there’s no way you can get high rankings by creating numerous pages to cover different keyword variations.
#10 How to optimize for RankBrain?
1. Utilize natural language
Basically, the one and only recommendation on RankBrain optimization we received so far came from Gary Illyes, Webmaster Trends Analyst at Google, who said that:
“Optimizing for RankBrain is actually super easy, and it is something we’ve probably been saying for 15 years now, is — and the recommendation is — to write in natural language. Try to write content that sounds human. If you try to write like a machine then RankBrain will just get confused and probably just pushes you back.
But if you have a content site, try to read out some of your articles or whatever you wrote, and ask people whether it sounds natural. If it sounds conversational, if it sounds like natural language that we would use in your day to day life, then sure, you are optimized for RankBrain. If it doesn’t, then you are ‘un-optimized. ”
I’m pretty sure this single official Google recommendation is of no surprise to you and you’re already creating content, targeting it at human readers in the first place. However, there are still a couple of things that also need to be taken into consideration while optimizing for RankBrain.
2. Find out search intent
As I’ve already mentioned, RankBrain’s ultimate task is to supply you with the most relevant search results possible. Therefore, it’s highly important that your pages match the desired search intent because it almost always entails high CTR.
So the first step towards optimizing for RankBrain is trying to understand the search intent behind your keywords. Simply type your keywords in the search box, then have a look at the results that Google comes up with, and try to figure out the search intents behind your keywords.
Here’s, for instance, what Google comes up with when you type in “green smoothie”:
The whole first page of search results is green smoothies recipes — Google doesn’t even bother to give you a definition of “a green smoothie”, so it’s safe to say that the search intent behind the query is “how to make a green smoothie”.
And when you search for “champions league”, Google assumes that you want to know about the results of the recent matches and delivers scores, players who scored, and a whole bunch of other stats. What’s more, when you start typing “champions league”, Google informs you about the score and date of the latest match directly in the search bar.
3. Improve relevance and comprehensiveness
Chances are that some of your pages don’t really match the desired search intent. If that’s the case, you need to work on improving your pages’ relevance and comprehensiveness.
As I already mentioned, with RankBrain shifting focus from keywords to topics, you need to make your content pages as comprehensive as possible by diversifying them with related terms and synonyms. What is more, with RankBrain’s ability to process natural language, try to refrain from unnatural phrasing, especially in titles and meta descriptions.
That said, the best way to improve your content’s relevance and comprehensiveness is by utilizing Rank Tracker’s Competition TF-IDF Explorer, which lets you collect tons of relevant terms used by your top competitors.
- Just open your project in Rank Tracker and move to Keyword Research > Competition TF-IDF Explorer.
- Then type in your keywords and wait for the tool analyse 10 of your top competitors and collect keywords that they have in common.
4. Check your snippets
Knowing that CTR is one of the things that RankBrain takes into consideration when estimating page relevance, it’s dead important to make sure that your snippets are well-optimized as they directly influence CTR.
Consider using Google Search Console to spot pages with low CTR. After that, have a look at their snippets and see how they can be improved.
5. Continue working on improving your rankings
Even with personalization of search getting bigger, the value of rankings is still high. In the RankBrain era, when users search for your target keywords and your site appears among top results, it becomes their preferred entity. It means that your website has very high chances of ranking high for similar subsequent searches of your competitors. Putting it simple, the higher your rankings are, the more likely you are to rank for similar queries.
6. Monitor your niche
The last but not least thing I can highly recommend you do on a regular basis is keeping an eye on your niche. The thing with RankBrain is that it can re-adjust SERPs if it thinks the search intent of your keywords has changed. So if Netflix launches a TV series with a name similar to your brand’s, reviews of the series will outrank your pages. If you don’t want this to happen, make sure to constantly monitor SERPs for your keywords. The easiest way to do it with the help of Rank Tracker’s SERP History tool, which lets you immediately spot important changes in the SERPs for your every keyword during each of your ranking checks.
The thing we know for sure about Google’s ever-changing algorithms is that they never stay the same for long. And chances are extremely high that Google will be (if not already) tweaking and refining RankBrain. What’s more, RankBrain is constantly learning and changing, so the only right thing to do is targeting your content at humans, keeping it relevant to search intent as well as making it conversational and up-to-date.
Just as usual, I’m dying to hear back from you. Please share your thoughts about RankBrain in the comments — see you there!