Google Rankbrain – A Guide on Google’s AI and its impact on search

RankBrain is a relatively recent machine learning algorithm that helps sort Google’s search results. Also, it allows Google to learn more about search queries and understand them on a deeper level. It is a model designed for query interpretation.

The news of RankBrain was released through an article by Bloomberg in October 2015. The piece discusses a key reason why Google opted to utilise AI-lead web search: RankBrain was able to outperform Google engineers by selecting the best search result more than 10% more effectively. So, in short, Google asked its engineers to select the most relevant search result for a given query. Then, it asked RankBrain to perform the same task. 70% of the time Google engineers were able to identify the best page, while RankBrain chose the best option 80% of the time.

Google’s application of AI to web search will come as no surprise to people who are familiar with machine learning – it’s an efficient way of consistently improving search results and user satisfaction. Before RankBrain, Google engineers hand-coded every element of search. Now, RankBrain can make tweaks to the search algorithm and add to Google’s knowledge bank independently. We’ll talk more about machine learning and AI further into this guide and go into detail about why you should understand what each of these terms means.

What does RankBrain do?

Here are the primary, simplified elements of what RankBrain does.

  • Helps Google make educated guesses about search queries that have never been searched before, which is about 15% of all queries.
  • Analyzes user satisfaction of search results and modifies the search algorithm based on that data.

Put into more detail, first, RankBrain takes new searches that Google would generally have little idea of what to return and uses what it knows about how each word is used contextually to suggest some possibilities for searcher intent. Like we stated before, about 15% of all search queries are phrases that Google has never seen before. This seems unlikely, but it’s true. Also, when users do ambiguous searches, RankBrain helps Google decide what the user is most likely looking for.

For example, searching for “Pokémon game” is somewhat ambiguous. What is the searcher intent? Is the user looking for the most recent Pokémon game or the most popular one? Are they looking to read about Pokémon games in general? Maybe the user is searching for information about an upcoming game release.

pokemon game Google SearchYou might be able to infer from this example that using a simple search algorithm simply won’t cut it for new or ambiguous searches. Search results are unlikely to perform well and users will probably get frustrated and give up when they don’t find the information they’re looking for quickly. A search algorithm that ranks based on typical ranking signals like backlink profile, content quality, and length of content won’t consider the fact that search results may need to be more varied in this kind of situation. These searches happen a lot. RankBrain is allowing Google to do a better job at comprehending the intent of the searcher and at delivering ideal search results.

The second way that RankBrain helps improve search is by taking search results with a low user satisfaction rating and modifying the algorithm for the keyword. It will increase or decrease how different ranking factors (like social shares, backlinks, and content freshness) influence the rank of each page. Then, RankBrain will analyse user satisfaction with the modified search algorithm. If it’s successful, the changes will be applied. If not, RankBrain will not alter the algorithm. Through this procedure, RankBrain can create an ideal algorithm for the selected keyword.

What is machine learning (ML) and artificial intelligence (AI)?

Before we dive too deep into RankBrain itself, it’s critical that you understand the basics of how machine learning and AI work. By understanding the technology behind RankBrain, you’ll be able to comprehend it in more depth than someone who hasn’t taken the time to teach themselves basic information about AI and machine learning.

Artificial intelligence describes a “smart” computer system that was created with the purpose of learning from its environment and maximizing successful results. The computer essentially mimics brain function and can gather, organise, and analyse data to achieve results that are or will eventually become better than human-driven results. So, the basic definition of AI isn’t too technical at surface-level.

Machine learning is a little more specific and more relevant to the way RankBrain functions. Machine learning refers to a type of artificial intelligence that allows machines and systems to improve its performance consistently without new programming by learning from its experience.

While reading articles about RankBrain, you’ll find that most websites use the terms “machine learning” and “artificial intelligence” interchangeably. This isn’t exactly accurate. If you understand the definition of machine learning, you understand exactly what it is that allows RankBrain to learn, evolve, and improve search results.

How does RankBrain work?

As we mentioned before, 15% of all searches are new and Google has no data for them regarding searcher intent and user satisfaction. This equates to around 500 million searches per day, so it was a huge problem for Google. Before implementing RankBrain, Google would return pages with the exact keyword phrase or the closest results they could find as “best guess” results. This basic way of handling unknown searches resulted in unhelpful search results most of the time.

Now, RankBrain tries to understand what you mean, sort of like a human might. When faced with a new or ambiguous search query, it will return a more accurate, more relevant set of search results. Understand that RankBrain is constantly improving itself and modifying the search results it displays based on user satisfaction, so you’ll see new results for the same query quite often.

RankBrain can access some databases that are based on entities (people, places, things, ideas). The original information within these datasets served as a foundation to establish the machine learning process. When someone performs a search query in Google, the query is broken down into “word vectors”, and each representation has a specific assigned address. Similar words “live” near each other.

Correlation between datasets using Google Bigquery
Correlation between datasets using Google Bigquery

When RankBrain runs into an unknown query, it makes use of the distance between vectors to guess multiple closely-related results for the user. As time goes on, RankBrain’s ability to return more relevant results is improved because of the data derived from the user experience.

How does RankBrain determine the user satisfaction of search results?

So, how in the world does RankBrain decide if search results are meeting the needs of users or not? Quite a few signals come into play here.

After a search is made, RankBrain shows you a set of results that it believes will be of use to you. If a lot of people end up clicking on one particular result, it will boost that page because it’s helpful to more people than the other options. If some of the pages it returns for a search query don’t get clicked on much, it will push those results down farther and replace them with other content.

RankBrain is actively looking at the following UX signals to determine user satisfaction: The click-through-rate of each page, the amount of time spent on each page (or the “dwell time”), the bounce rate, and “pogo-sticking”. How users interact with the search results and the pages they click on helps RankBrain decide whether or not the content is valuable and meeting the needs of the users.

You likely know how to analyze your page’s organic CTR, bounce rate, and dwell time (use Google Analytics). Let’s quickly illustrate the concept of pogo-sticking.

Let’s say that you type “iPhone photography” into Google. You’ll probably click on the first search result – this is what most people do.

Annoyingly, that first article isn’t helpful because it’s short and generalizes to the extent of being relatively pointless.

Now, you click the “back” button to return to the SERP and locate the second search result.

Search result number two proves to be similarly unhelpful – it’s long, but full of filler content and doesn’t provide an in-depth discussion of the topic. Onto the next article.

The third result is almost exactly what you were looking for. It even expands on several related topics and teaches you few new facts about photography in general. You spend eight minutes reading the initial article you clicked on and then spend a further ten minutes browsing the rest of the site. After that, you didn’t return to the SERP because you found the information you needed.

The movement back-and-forth between the SERP and each result is something RankBrain takes into account heavily. When you click on a result and quickly press the “back” button or leave by another means to click on a second page, it tells RankBrain that the first page wasn’t a high-quality result. When RankBrain finds that people typically stop pogo-sticking on a certain search result, they will infer that it’s better content and will boost it in the search.

How to optimise for RankBrain?

In some ways, you can’t “optimise for RankBrain” but in some ways, you can. What you can do is optimise for specific ranking signals that are more influential in the search algorithm for a particular keyword. Becoming skilled enough to identify what signs are most important for each keyword will take some serious time and effort, but you’ll be able to improve your rankings more effectively.

Regardless, it will still benefit you and your team to understand the way Rank Brain modifies search results. If you’re able to identify ranking signals that RankBrain determines to be of high-importance for your keywords, you’ll have an advantage because you’ll be able to optimise for these specific signals to a certain extent.

One of the most important takeaways for SEO is that with Rank Brain, the context of your keyword matters. As expert Bill Slawski said,

“To an equestrian a horse is a large 4 legged animal, to a carpenter, a horse has 4 legs, but it doesn’t live in fields or chew hay, to a gymnast a horse is something I believe you do vaults upon; with RankBrain context matters, and making sure you capture that context is possibly a key to optimizing for this machine learning approach.”

This quote clearly illustrates why Rank Brain is an essential element of search and why establishing the context within your content is paramount.

In some ways, “optimizing” for RankBrain is more complicated than optimizing webpage before RankBrain. Since ranking factors can vary in terms of importance depending on the search query, there are a lot more moving parts to the Google search algorithm than during the pre-RankBrain era.

Still, in some ways, optimizing for RankBrain is less complicated because RankBrain relies heavily on the context of your keyword use and the quality of your content. Content has become more and more central to SEO within recent years, so many SEOs won’t find themselves reworking their SEO strategy entirely. If your SEO strategy is less content-centric, however, you may find yourself having to make changes at the core of your technique.

Here’s what you can do to “optimize for RankBrain”.

Match results with search intent

Sometimes, you can predict what ranking factors that RankBrain will allow to have more of an effect on a specific keyword. Do a search for your chosen keyword and analyze the results that are returned. Check out the link profile of each website. How many social shares does each piece have? What kind of traffic does the website get? Have you heard of the website before? When was this content published? Make a note of the specifics for each ranking factor and you’ll get a sense of what factors are more valuable for this query.

To give you a better idea of how varied the influence of different ranking factors may be, here is an example.

Consider search queries about breaking news: Natural disasters, amber alerts, and election results, for example. The relevance of results for the search query “volcanic eruption” will likely have more to do with how fresh the content is than how many links the piece has accumulated. In contrast, consider a search for “History of classical music”. In this situation, the depth and authority of the content are more relevant.

Stop with the one-kw-one-page tactic

It’s highly likely that you’ve already stopped using this strategy but if you haven’t, now is the time to change. Creating different pages for similar keywords used to be a popular SEO practice, because each keyword had separate search results. This strategy is obsolete because of Hummingbird and RankBrain.

The reasoning for this is simple – RankBrain helps Google better understand content contextually. So, Google can now see that the phrase “best yoga moves for beginners” is essentially the same thing as “best beginner yoga positions”. Google understands that the searcher intent for both keywords is the same, thanks to RankBrain. The SERP for both of these pages likely looks very similar, while several years ago they would have focused on results with an exact keyword match.

Instead of creating pages with slight keyword variations, put together a page with valuable, long-form content that includes enough LSI keywords, synonyms, and keyword variations for RankBrain to understand the main idea of the content and determine what queries it will be relevant to.

Pay attention to time on site

We’ve talked a lot about how RankBrain utilizes the user satisfaction of search results to decide which results are high-quality and which results are not-so-stellar. Monitoring and improving the average amount of time users spend on your website can help you out in this area.

Avg Session duration

Use Google Analytics to get a picture of your current average dwell time. Depending on your industry or niche, the ideal number will vary greatly but shoot for getting it as high as possible. Seven minutes or more is reasonably outstanding. Less than two minutes signifies you need to make improvements. Here are some of the things you can do to improve your dwell time.

  • Use internal links intelligently. When you can direct users to more of your content, your dwell time increases.
  • Get to the point. Cut down your article introductions if they are longer than five or six sentences. Readers click on your content for the information they are promised in the title. If you don’t deliver quickly, they are more likely to lose interest.
  • Refrain from placing content below-the-fold. If the first thing your site visitors see when they click on your title is a large header photo, they will be more likely to return to the SERP to find another page. Don’t make them have to scroll down or you will lose viewers. Place all of your content above-the-fold.
  • Remove excessive advertising. Ads can easily detract from the user experience. If your website looks cluttered or spammy, people are more likely to leave.
  • Break up your content for easier reading. Use subheaders religiously and try not to end up with paragraphs longer than six to eight sentences. Depending on your niche, you may want to break down your paragraphs even smaller. Reading long-form content is much easier when it’s broken down into digestible chunks – a wall of text won’t look as enticing to your visitors.
  • Reassess your content and make sure that is it all or most of the following: Useful, unique, entertaining, informative, easy to skim, and well-designed. Identify any weak points and make the necessary changes.

Build up a website’s overall reputation

Improving the reputation of your website will always help you rank higher more easily. Study up on how to get more social media shares and be interactive with your followers. Start creating video content and measure the results of your SEO strategy. Creating specific, actionable goals, being consistent, and measuring your progress is the only way to move forward and gain traction online.

Your site will benefit greatly from more brand awareness. Just think about it for a few seconds – if someone performs a search and scans over the results, they’ll be more likely to click on your site if they are already familiar with your name. Building brand awareness should be at the top of your to-do list every day, no matter what size following you have online.

You can use Facebook, Twitter, and Instagram ads to build brand awareness. Additionally, publish guest posts on blogs with high traffic and become a guest on some podcasts to get your name out there. Create multiple opt-ins for your email list and use A/B testing to find out what your best performers are. Once they are subscribed, give them an email newsletter or another resource so valuable that they don’t want to unsubscribe from you. Only send your subscribers the best of the best – you want your brand to be associated with the high-quality content.

Explore influencer marketing opportunities to consistently gain exposure to new audiences. Even smaller businesses can usually afford to work with micro-influencers (influencers with under 50,000 followers). This marketing strategy is ultra-targeted and micro-influencers boast incredible engagement rates.

Focus on medium-tail keywords

Since creating a page for “best business strategies for 2018” and another one for “2018 best business strategies” is no longer a viable SEO technique, optimizing long-tail keywords are becoming less and less valuable.

Utilize higher-volume, medium-tail keywords. Medium tails aren’t as competitive as short-tail keywords because they include two to three words. When you successfully optimise amazing content around a carefully-selected medium tail keyword, RankBrain will automatically identify hundreds of similar terms to rank you for, in addition to your targeted keyword. This is because RankBrain understands the correlation between “top blogging tips” and “tips for bloggers”.

One well-optimized page with high-quality content can rank for more than a thousand keywords simply by optimising around one medium-tail keyword.

Improve your CTR

RankBrain loves content with a high organic click-through-rate. Get more eyes on your content by learning to write titles and meta descriptions that stand out. Here are a few ways you can accomplish this.

Use emotionally-loaded wording

Using words in your titles that convey a strong emotion (whether positive or negative) will increase your CTR and social shares. Generally speaking, headlines that convey a strong positive emotion do better than negatively-loaded headlines. That being said, there are some cases where negative emotions are more appropriate for the title. The main trick is to use powerful wording without sounding spammy or like an obvious advertisement.

Try to craft titles with a high EMV (Emotional Marketing Value) score. The higher the score, the more likely your title is to have an effect on the reader and garner clicks and shares. You can test your titles for their EMV score by using the Emotional Marketing Value Headline Analyzer from Advanced Marketing Institute, which is a free tool that is based on the research conducted by the institute. Ideally, you’ll want your headlines at an EMV score of 30-40% or above. Your EMV score is calculated by analyzing the percentage of your title made up of EMV words.

Not only will this tool provide you with the EMV score for your title; it will categorize it as either intellectual, empathetic, or spiritual. Each of these categories is ideal for a different kind of piece, and you should get to know them if content marketing is a part of your overall marketing strategy.

If you don’t already use BuzzSumo to brainstorm ideas for your content and headlines, you should start now. In short, it will show you fresh content in your industry or niche that is performing the best. If you can create content and titles that are on-par with or better than some of what you find using BuzzSumo, you’ve got a recipe for success.

Include numbers and top-performing headline trigrams

Try to publish list posts regularly and explore popular trigrams to use in your titles. A trigram is a three-word phrase, and BuzzSumo has some interesting data about which phrases garner the most engagement. “Will make you”, “this is why”, and “can we guess” are all top choices to use in your titles.

Buzzfeed, possibly the king of list posts, has a global online presence, publishes thousands of “listicles” every year. Their top-performing list posts are lists of 5, 10, 15, and, for some reason, the number 23 is a consistently high-performer. Why is BuzzFeed’s content so list-heavy? List posts get more social shares than most other content formats. Not only are they popular, but list posts are also simple to put together and tend to provide the reader with lots of value.

Understand that reader’s spam-filters have evolved

When your titles sound spammy, they are less likely to get clicked on. Phrases like “magic”, “easy steps”, and “simple tricks” are terms that many users have been pelted with left and right over the years. Also, don’t go overboard with trying to create titles that instil a sense of urgency. As a marketer, you probably know a lot about creating urgency. Using words like “need” and “limited time” in your titles can hurt your CTR significantly. Users are exposed to so many emails, articles, and videos that are created to feel urgent that these buzzwords no longer have the same effect.

Use brackets or parentheses

For whatever reason, titles that include brackets or parentheses do about 38% better than titles without either of these elements. Instead of writing “How to Become a Vegan”, try “How to Become a Vegan (Recipes, Substitutions, and Health)”. Providing additional information about your content without getting too wordy gives users a better idea of what they can expect. You can include “[Interview]”, “(2018)”, or [Case Study] after your title if they are relevant.

seo 2018 Google SearchInclude “power words.”

This subset of specific words packs an impressive emotional punch and are regularly used in high-performing article and video titles. Words like “ultimate”, “the truth about”, “proven”, “now”, and “surefire” are simple and powerful. Do some more research about power words and start including them in your titles. Try referencing this helpful list of 180 power words that CoSchedule put together.

Content is still king

At the end of the day, we’re left with several new things to consider when optimizing our websites and pages for RankBrain, but it’s clear that crafting compelling and shareable content is still topping the list of importance for SEOs. Gone are the days of keyword stuffing and ranking a separate page for each long-tail keyword you target. Today, it’s all about how users interact with your pages and how helpful Google determines them to be, with the help of RankBrain.

Improve and update your content with these steps.

  • Determine what your goals are and be specific.
  • Rework your keyword research strategy to work around optimizing for medium-tail keywords.
  • Get familiar with CoSchedule’s Headline Analyzer, and Advanced Marketing Institute’s Headline Analyzer.
  • Create an editorial calendar and stick with it.
  • Develop more long-form content and use internal linking strategically.
  • Include pictures and video content whenever possible. Don’t forget to optimise images to ensure faster loading times.

RankBrain FAQs

Is RankBrain the new Google search algorithm?

No, it’s just part of Google’s search algorithm. RankBrain is considered one of the hundreds of ranking factors, although it is near the top of the list of importance.

How many ranking signals does Google use?

While no one knows the exact number of Google ranking signals, (except Google) we can make educated guesses with the information we do have. There are likely several hundred major players. Google stated in 2010 that there are over 200 ranking signals. New, the verified information we’re able to glean from Google about ranking signals occasionally makes an appearance. Google’s Webmaster Guidelines outlines some major ranking factors. Speaking at PubCon 2010, Matt Cutts of Google implied that there might be as many as 50 sub-signals for each of the 200 major ranking signals, leaving us with up to 10,000 factors that come into play within the Google search algorithm.

Does RankBrain only apply for unknown search queries?

No. While RankBrain was probably created because of the fact that Google needed a better solution for their growing number of unknown search queries, RankBrain’s AI capabilities come in handy for other search queries. In any situation where the user satisfaction of search results is not ideal or starting to dwindle, RankBrain may come into play.

In summary

There’s a lot you can do to boost your rankings in a RankBrain world. When in doubt, publishing lengthy and valuable content and building brand awareness will never do you wrong. Optimize your content around a medium-tail keyword and make a point to sprinkle LSI keywords and internal links throughout your posts. Do not ignore your dwell time, bounce rate, or exit pages. Use these key performance indicators to hone in on your site’s weaknesses.

Thanks to machine learning technology, RankBrain is getting more accurate each day, so don’t assume that you can get away with outdated SEO strategies for much longer. Use this guide to make a plan and start optimizing your site for RankBrain.