In this post, I’ll attempt to give you a point of departure on your journey to understand how RankBrain (and AI in general) will ultimately impact organic search and, more importantly, SEO.
So, without further adieu, let’s get into it…
What is RankBrain?
RankBrain is now Google’s 3rd-most-important ranking signal and, as I mentioned, a machine-learning, artificial intelligence system that is designed to process, sort and deliver more relevant search results to user queries. The system was first confirmed to be in use by Google back in October 2015, and has been gaining momentum ever since. The immediate impact of RankBrain, upon its introduction, was to change the rankings of over 30 trillion different webpages, instantly becoming the 3rd most significant factor in ranking pages (after content and backlinks).
Through the application of RankBrain, leveraging artificial intelligence and machine learning, Google gathers knowledge and makes new connections, both from what it’s taught AND, more radically, from what it teaches itself. RankBrain, in the simplest terms, is a system designed to sift through billions of webpages looking for those most relevant to queries Google can’t understand, which, prior to RankBrain, constituted a very large portion of the long-tail queries it saw everyday.
Why was RankBrain introduced?
Prior to RankBrain, Google had a big problem understanding certain types of queries; specifically those, as mentioned, that fell in the long-tail category and were categorized by Google as “unique or ambiguous”. These query types had historically stumped Google and thus RankBrain was developed to tackle that problem.
In the days before RankBrain, Google utilized a basic algorithm to determine which search results to show for all types of queries, including those aforementioned ones deemed “unique or ambiguous”. For example, if your search comprised: “car accident settlement in houston,” it’d scan it’s index to determine which results had those exact keywords. That is, the system wouldn’t try to comprehend the intent of the query, instead, it’d simply look for pages with those words: “car”, “accident”, “settlement”, “houston”. A good system for the most part but one fraught with the potential to miss the target more often than Google (or users) wanted. In fact, at the time, about 15% of all search queries Google saw were brand new to it, meaning, they had never seen them before and therefore had no clue about what the searcher really wanted to know. So what’d it do? Well, basically just guess. And 15% of total Google searches per day equates to a mind bogglingly high 400 million that the system just couldn’t understand and, consequently, would deliver against sub-optimally. Hence, RankBrain was created to solve that problem.
Today, employing RankBrain’s artificial intelligence, Google is focusing on better deciphering intent behind a query before returning search results against it and then gauging user satisfaction with the results given after delivery.
A few ways it accomplishes those goals are:
- To Correlate New Queries to Old Queries
RankBrain draws correlations between never-before-seen queries and those it has seen before to better understand the intent of the former. The thinking is this: if Google already knows the intent of the latter (a query it HAS seen) and can match it up with the former (a query it HAS NOT seen) then by transitive property it might draw assumptions about the intent of the former. Confused yet? Example…let’s say RankBrain knows that most people who search for “car accident settlement in houston” are looking for attorneys that specialize in car accidents, versus car accident settlement stories, AND it knows that “I had a car crash outside south houston what should I do?” correlates to the original query (“car accident settlement in houston”) with respect to searcher intent, then it will return results similar to the ones it returns for the original query.
- To Understand Concepts Beyond Just Words
Concept search is a way of matching queries and search results based on concepts versus the old way of matching them based on keywords. The old way (aka Keyword-Based Search) had significant limitations when dealing with large text documents due, in large part, to the inherent ambiguity in language (two examples of that ambiguity are: “synonyms”, or two or more words with the same meaning, and “polysemes”, or single words with multiple meanings) and consequently, as the universe of large text documents out there online grew and grew, search results, based on traditional keyword ranking methods, became ever more rife with non-relevant items. Concept Search attempts to solve this problem instead by conceptually matching query and search result. In other words, by first understanding the “idea” expressed in the query (the “concept”) and then looking to match it with webpages expressing the same (or a similar) idea. So in essence, what RankBrain is designed to do is to turn a search query into a concept and then try to find webpages that cover that concept so it can return them in search results in response to that query.
- To Take Into Account What It Knows About The Searcher
RankBrain also factors in aspects about the searcher it might know, or might divine from their behavior, such as, location and search history, as well as, how someone actually interacts with search results. With respect to the latter, RankBrain does that by taking note of so called “user experience signals (UX signals)”, for instance, whether or not searchers are clicking on search results, which ones they are clicking on and how long they’re staying at the destination page after clicking through. UX signals are strong indicators of the level of satisfaction users are getting from the search results and Google uses them for ongoing refinement to its search rankings. For example, if I click on one of the items in the search results list but quickly hit the back button once I get there, and then choose another item from the same results list and, this time, end up staying on that page for ten minutes it’s probably a good indication that: A) the first result did not satisfy my need and B) the second one did and did so pretty well. And if that kind of user behavior happens often enough, it probably indicates that the first result is a poor fit for the query. Of course, with enough data, you could also draw other conclusions, like that the content on the first result is so poor that it wasn’t a matter of it not delivering on the underlying need of that query but rather that, it might not suffice for ANY query.
Understanding how RankBrain Operates
Now to further help understand the workings of RankBrain, let’s say Google is faced with the query: “Summer Olympics location”, it might possibly interpret it in one of the following ways:
- The searcher is asking for information about the host country of the next summer games.
- Or, alternatively, the one where it took place last.
- Or, information about the actual games.
- Or, directions to the myriad venues where events and games are taking place.
- Or, maybe even, historical information about the very first Olympic games.
All these derivatives are easy for the human brain to deftly. Yet, for the simple search algorithm, that’s a tall order. In formulating a response, the simple search algorithm would factor in, for instance, relevance of keywords and the quality of backlinks candidate pages have, and likely only return results about the last Summer Olympics in Brazil. Why? Well because content related to it is so abundant and has earned millions upon millions of backlinks since published. In this scenario, it’s very unlikely that the location of the next Summer Olympics in Tokyo would even be considered. In short, this is the gap in understanding that RankBrain was developed to bridge.
RankBrain was designed to emulate the learning style of the human brain including the identification of patterns in search behavior. For example, if the majority of people searching for “Summer Olympics location” are looking to find out where the next olympiad will be held, the search engine will adjust to prioritize information in line with that intent.
Additionally, leveraging what it knows about a searcher’s current location, and factoring in the date and time, the system would adjust results. For example, if a searcher is in Tokyo at the time of the games, and types in that same query, they might very well get driving directions to the sporting venues, highest in results. These advancements in their search algorithm, help Google provide results that are more likely to satisfy searchers.
What are the SEO implications of RankBrain?
From a search engine optimization perspective, RankBrain cannot be optimized for, at least not in the same way we’re used to with Google’s older algorithms. So what to do? Well my best guess—since Google has been quite tight-lipped about it—is that, since RankBrain helps classify and sort pages based on their content, site owners should concentrate on developing deeper, more comprehensive, content in their topical area of focus.
To that end, here are some suggestions for SEO.
- Use Natural Language
Gary IIIyes from Google said that one way for site owners to optimize for RankBrain is through writing content in natural language. But isn’t that what we should always be doing anyway? Well yes, but people often don’t. Writing in natural language means making it sound human versus written for a search engine (i.e. jammed pack with keywords that detract from the natural flow). Make sure your content is quality, easy to read, and sounds natural, otherwise RankBrain will ding it. You CAN still optimize for keywords just as long as it sounds natural. Read your content aloud. It helps you recognize those parts that don’t.
- Consider The Benefit To The User
Before the advent of RankBrain, traditional methods, such as keyword seeding, played a huge role in search optimization. Today, best practices are to be more vigilant and mindful for how your content will serve the interests of users. For example, when optimizing for a subject like an impending hurricane, focus attention on content freshness since, given the topic, newsworthiness is most valuable. Users searching for info on a hurricane are, more than likely, looking for current storm info. Whereas, if your topic is “The American Revolution”, your focus and priority would need to more on content depth and authority (backlinks). As a proponent of user-centered digital marketing, I find the fact that Google has had to essentially legislate this absurd. Yet, they have and it was necessary.
The bottomline is: RankBrain will continue to become more and more sophisticated given its ability to learn, and, might ultimately overtake links and on-site factors as the primary ranking signals. So all site owners should heed this trend, get educated and focus on delivering more value to their visitors.