For nearly four years now, Google’s core algorithm has benefited from a brain . . . of sorts. RankBrain is a signal that employs machine learning to influence how the search engine produces results pages. It was designed to learn new things over time, build on its knowledge, and make valuable connections to improve the search experience for users. Since it was first revealed in October 2015, we’ve been trying to learn more about how RankBrain works and how we might take advantage of any insights. We’re happy to share what we know!
How RankBrain Works
When asked how RankBrain works in an Ask Me Anything (AMA) on Reddit, Gary Illyes, a Google webmaster trends analyst, explained it this way:
RankBrain is a PR-sexy machine learning ranking component that uses search data to predict what would a user most likely click on for a previously unseen query. It is a really cool piece of engineering that saved our butts countless times whenever traditional algos were like, e.g. “oh look a “not” in the query string! let’s ignore the hell out of it!”, but it’s generally just relying on (sometimes) months old data about what happened on the results page itself, not on the landing page.
Greg Corrado, a senior research scientist at Google, explained RankBrain’s function like this: “If RankBrain sees a word or phrase it isn’t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.”
Essentially, RankBrain uses its knowledge of search data to predict the intent of searchers when they enter a search query that’s never been used on Google before (about 15 percent of searches are new). Unlike traditional algorithms that sometimes got stuck on new queries, RankBrain’s AI-like skills allow it to solve new problems with finesse.
Does RankBrain Use On-Page Signals?
Some SEOs have speculated that RankBrain uses on-page signals as well, but Illyes shot down these theories at the end of his explanation: “Dwell time, CTR, whatever Fishkin’s new theory is, those are generally made up crap. Search is much more simple than people think.” The jab at Fishkin refers to Rand Fishkin, a co-founder of Moz.com, who has claimed that RankBrain uses these metrics.
According to Illyes, RankBrain gathers data from searchers’ interactions with search results, not from their interactions with the content itself.
Understanding the True Intent of a Query
RankBrain’s goal is to discern the true intent of each new query. Sometimes this involves deciphering queries that include misspellings, typos, omissions, unusual phrasing, synonyms, or the wrong word being used. Other times, RankBrain must understand that the searcher has included a negation (i.e., no, not, never, without). Or, the user may be searching for something that just happened, and RankBrain must identify the best results available for this brand-new request. Whenever Google receives an original query, the algorithm uses RankBrain to figure out what the user really wants to know. RankBrain identifies the existing search results that best suit the query.
Wondering how RankBrain works? According to Corrado, this is the process:
“. . . RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand. If RankBrain sees a word or phrase it isn’t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.”
So unlike a human brain (which can analyze a sentence into parts and understand semantics) or a Natural Language Processor (which uses sentence structure and linguistics to understand intent), RankBrain translates words into numbers and uses a database in order to understand them.
The system is based on entities, which are things and concepts that are singular, well-defined, and distinguishable. Each entity receives a unique identifier and is plotted on a chart so that RankBrain can better comprehend the relationships between the entities. Similar word vectors share similar “addresses” on the chart. Using its growing database of entities and their connections, RankBrain is able to make sense of unknown queries. Even if a result isn’t optimal, that data is fed into the database to (hopefully) produce more relevant results the next time.
While it’s fun to talk about RankBrain truly understanding searchers and their queries, it’s important to remember that RankBrain is still a computer program that relies on mathematically mapped out identifiers, not a human brain with a comprehension of linguistics.
The Difference Between RankBrain and Neural Matching
Last year Google revealed that 30 percent of all queries were impacted by a system known as neural matching. Neural matching is “an AI-based system Google began using in 2018 primarily to understand how words are related to concepts . . . It’s like a super synonym system.”
Google knows that people sometimes don’t know how to perfectly word their query. Or, they may only have a vague idea of the things they’re trying to identify. The example they gave is that someone might search, “Why does my TV look strange?” The first result Google produces is about the “soap opera effect,” which is a feature of some modern televisions caused by a perceived increase in frame rate due to motion interpolation. This odd phenomenon may very well be what the searcher was trying to identify, though he or she didn’t have the words for it. Google is able to figure this out through neural matching, not RankBrain.
According to Google, RankBrain helps Google relate pages to concepts, while neural matching helps relate words to searches. So in a way, RankBrain is working to understand the intent of a page, while neural matching is working to understand the intent of the searcher.
Optimizing for RankBrain
Now that you better understand how RankBrain works, you’re probably wondering how you can fine-tune your content to make the most of the program. According to Illyes, “Optimizing for RankBrain is actually super easy.” The key is to write in natural language. What does that mean? You need to write like a human, which should be effortless . . . We’re all humans here, right? Try to write in a style that sounds conversational, in language you would use in your everyday life.
While most people will find this style of writing easy, it may be tricky if you’re not a native speaker of the language. If you aren’t sure if your writing sounds natural, ask a coworker or friend what they think.
Finally, it’s important to remember that RankBrain is always learning and changing, automatically improving itself based on past experiences. Trying to optimize for a moving target is pretty tricky, so we encourage you to focus on creating high-quality, relevant content. What else is new?
Hoping to get a leg up with Google and other search engines? Check out 417 Marketing, an online marketing company based in Springfield, Missouri, that specializes in SEO and web design. Click here to contact us and learn more about what we can do for your company.