Google’s RankBrain and Search Engine Optimization (SEO)

Google’s search results seem to constantly fluctuate, and this is because they are! But why is that?

There isn’t one answer to that question, but a big part of the equation is how it interprets and contextualizes searches. And a big part of how it does that is its RankBrain system.

We’ve looked at the different subsystems in Google’s environment before. RankBrain, though, is among the most prominent and applicable to the largest number of searches. So it’s worth a deeper dive on how it works, and what it means for your SEO efforts.

Understanding RankBrain on its own can produce enormous gains for your website’s digital presence. So let’s start with the basics and build up to how you can leverage your expertise and unique voice to get the most out of search results, gathering leads and revenue for your brand!

What is RankBrain?

RankBrain is an AI (artificial intelligence) system within Google’s search algorithm environment that analyzes and attempts to understand search queries.

I call Google’s search an environment instead of a single algorithm for a reason, though. And that’s because there isn’t one master algorithm that analyzes every search query and delivers results. The many types of searches and content necessitate different subsystems that form parts of the algorithm. Depending on what you’re searching for, you may only be interfacing with a portion of its subsystems.

RankBrain is among the most ubiquitous subsystems, though, because it relates to the interpretation of search queries. Even as new AI systems begin to populate Google’s results with AI-generated answers, RankBrain’s underlying processes are still at work.

Search Intent

Possibly without realizing it, you’ve likely learned to phrase search results like you’d speak or type to a friend. But if a search engine interprets your words literally, this can sometimes produce problems.

A simple example that Google itself gives is misspelling a word. Are you actually searching for “piziza” near you? No, you’re looking for pizza.

You and I understand that immediately, but a logical computer engine doesn’t automatically understand the intent unless it’s trained to do so.

By creating the RankBrain system to be AI-based, it’s capable of improving itself over time by analyzing how people interact with search results and adjusting its understanding of search queries accordingly.

How AI Analyzes Search Intent

Misspellings are perhaps the easiest type of common interpretation that search engines are asked to parse.

But what about more difficult examples? What if you don’t remember the name of a movie but remember the plot of it? Can a computer algorithm help find what you’re looking for?

Yes, it turns out. It’s not always perfect, but RankBrain is extremely good at sussing out search intent even for very strange searches.

Keyword Matching

It’s tempting to think that a word-for-word match might be the best one, but this often isn’t the case. In our searches, we omit words that could be important to a topic, or sometimes use others that are less precise.

However, some keyword matching does occur in search results.

Google is even capable of toggling this. Have you ever done a search and it omits one of the words from search results? This often occurs with articles and prepositions. Your results page will often then give you the option to make that word mandatory in the result, to help refine the search further.

More tricky is when the same word can have multiple meanings. Are you talking about an elephant trunk or a tree trunk? A river bank or a place to store your money? A baseball bat or an animal bat?

This is where concept matching plays a role.

Concept Matching

Keywords can tell some of the story, but every individual idea resides within a larger context. This is where concept-matching comes into play.

RankBrain analyzes the entirety of a page, not just individual keywords, to determine if your query matches the content of a page.

For example, consumer statistics related to economic models and consumers in a food chain hierarchy are vastly different concepts. You want to get results on whichever of the two you’re actually searching for. Perhaps typing “consumer” won’t produce the best results, but “consumer predator” will automatically give RankBrain more information about the related concepts that you’re searching for.

The word “predator” might even be contained in some results that contain economic statistics, but the concept matching system is still unlikely to pick them as good results because it’s analyzing the entire page and all the concepts contained there.

BERT Model

BERT is another subsystem that was introduced in 2019, and portions of its systems are in use today. It provides more sophisticated models for how combinations of words relate to one another.

This is more important than you might realize. It’s estimated that as much as 15% of daily search queries have never been made before! 15% might not sound like a lot, but it becomes staggering when you think about the billions of search results that search engines have to parse each year.

So if there aren’t preestablished models for understanding a brand new query, BERT has to determine intent.

Are you searching for information? Searching for a specific product or service? Looking for a physical store location to purchase something or engage in an activity? Or something else?

This relates to the types of information that will be delivered. Should Google show you a local map with nearby businesses, or online shops offering the same things? Or both? Each of those can be the correct answer depending on the search intent.

It’s also why if you type “coffee” you’re likely not going to get a bunch of results about the history of coffee. Instead, you’ll see local businesses that sell coffee, because BERT and RankBrain have learned that this is the most common intent for that style of search.

MUM Model - Information Models as Results

MUM (Multitask Unified Model) is another Google subsystem that focuses on information models, and analyzes (and produces) more than just text. It can analyze videos, pictures, and other forms of media to assess

MUM isn’t used quite as often because these different types of search results are not as common as text results. Increasingly, though, you’ve probably seen more video, product and image results when you search on Google.

MUM helps to deliver these results in the proper format. It’s a multimedia extension of the same ideas that inform the RankBrain and BERT systems.

Serving Search Results and User Experience Data

Concept matching and information analysis is all well and good, but how does Google decide what to deliver in its search results? And how can it know those are good results for each person?

This is where user experience data comes into play and where AI can operate at speeds that humans can’t. Every time you make a search, Google’s checking to see what results you click on, and also how long you stay on a page.

For search queries, it’s also checking to see if you type in a different search query. This can happen when results don’t match with your expectations. This is a sign that its interpretation and concept matching wasn’t quite right and may need adjustment.

Clickthrough and dwell time are two metrics that help to define the user experience in Google’s system. Ultimately, Google bases its rankings on usefulness to the user. Tracking actual user behavior is central to this.

RankBrain is analyzing billions of these interactions, so it can form a very clear picture of the usefulness of some websites. It again doesn’t mean search results will always be perfect (far from it), but that results can - and will - improve over time as more data rolls in.

RankBrain and Your SEO

Now comes the portion where we tell you what this all means for you. Because what use is all of this unless it brings you more traffic, leads and revenue?

SERPs

The systems we discussed above - particularly knowledge-based ones like MUM - will attempt to figure out what type of resource your page is, and also what type of business you are.

Making sure all of your webpages are structured properly and have relevant copy, pictures, and metadata can help them appear in the correct searches for the best audiences.

This means appearing in the most relevant SERPs, or Search Engine Results Pages. Our full rundown on SERPs talks in more details about the different types and how you can target them.

Keyword Research

If it wasn’t obvious from the article above, you want to be targeting keywords that are the most descriptive to your products, services and brand. This means sometimes expanding beyond single-word keywords (somewhat counterintuitively, a keyword can actually be a phrase).

There’s also a phrase called LSI, or Latent Semantic Indexing keywords. These are important to be aware of.

Basically, these are keywords that routinely appear alongside related concepts. If you are trying to target a particular keyword, you’ll also want several LSI keywords that will help Google’s content matching systems.

Google has said that LSI isn’t a parameter in their search, and it’s entirely possible that this is technically correct. However, when related questions and concepts inform Google’s “People Also Ask” SERP, autofill results, and purportedly inform systems like RankBrain, it may just be that they aren’t defining them precisely as LSIs. 

Stated differently, in practice, including LSI keywords can be useful both for search engines and for users consuming your content.

Writing for User Experience

The biggest thing you can do is produce content that you believe will be exciting and engaging to your audience.

Why? Because clickthroughs and time-on-page (and related metrics) are how Google decides what pages are most useful. And this means creating content for humans, not algorithms.

Yes, there’s some data analysis involved, but you can never lose sight of who you’re actually creating webpages, articles, videos, podcasts, and other media for. If it’s interesting to your audience, it will succeed. And if it isn’t, it won’t. Everything else is just giving yourself a better chance to reach a wider audience.

At Leadflask, we help companies craft a meaningful digital presence through a variety of media channels. If you’re interested in learning more about our work, let’s talk!