RankBrain Tweaks Weights of Google Ranking Signals On Its Own?

Chris CrumSearch

Share this Post


Google revealed RankBrain to be its third most important ranking factor last year, but there appears to be a good deal of confusion within the company about just how it works.

What are your thoughts about RankBrain? Do you see it as a good thing for Google? Share your thoughts.

RankBrain was revealed in October pretty much out of nowhere. It didn't come in an official announcement, but from an interview Bloomberg Business ran with Greg Corrado, a senior research scientist at Google. He said that Google had introduced the algorithm on a wide scale earlier in the year and that it quickly became the third most important signal out of hundreds in Google's ranking algorithm.

RankBrain uses artificial intelligence to put written language into mathematical entities (vectors) that computers can understand. If it sees a word or phrase that it doesn't know, the machine guesses what words or phrases might have similar meanings. It helps specifically with never-before-seen search queries. Apparently it's better at humans (even Googlers) at guessing which results Google would rank number one for various queries. It's the first ranking signal that actually learns on its own. Google has indicated in the past that turning RankBrain off would be as damaging to search results as turning off half of Wikipedia pages.

Barry Schwartz reported on a session at SMX West earlier this month in which Google's Paul Haahr, who is described as "a top engineer involved in core ranking," admitted that Google itself doesn't fully understand RankBrain.

Now, Schwartz is pointing to a Twitter conversation with Google's Gary Illyes and Moz's Rand Fishkin in which it's suggested that RankBrain can't use a new factor that wasn't previously in an algorithm, but it might adjust the weights of existing signals. Still, Illyes notes that while he's on the search quality team, he doesn't know everything about RankBrain.

Last week, Stone Temple Consulting released some research on RankBrain's effectiveness at improving Google search results. The data was gathered by comparing 500,000 search queries from both before and after RankBrain was implemented.

They found that Google improved results on 54.6% of queries that it previously misunderstood. Examples of words and phrases RankBrain handles better, according to the firm, include: what is, who is, where is, without, not, and convert.

Stone Temple's Eric Enge suggested that Google may use RankBrain to impact selection of featured snippet results, trigger the delivery of a map where there wasn't one shown before, and/or determine if the main impact of a given query would be an improved search results snippet.

"Predictably, one of the most common questions I get asked is how RankBrain will impact SEO," he said. "Truth be told, at the moment, there is not much impact at all. RankBrain will simply do a better job of matching user queries with your web pages, so you’d arguably be less dependent on having all the words from the user query on your page."

"In addition, you still need to do keyword research so that you can understand how to target a page to a major topic area (and what that major topic area is)," he added. "Understanding the preferred language of most users will always make sense, whether or not search engines exist. If you haven’t already (hopefully you have!), you can increase your emphasis on using truly natural language on your web pages."

According to Enge, the real impacts of RankBrain are an increase in overall search quality and in Google's confidence that they can use machine-leaning within the core search algorithm.

Be sure to check out Stone Temple's study if you haven't already. It includes a nice infographic outlining the highlights.

Do you think RankBrain is a positive thing for search results? Discuss.

Chris Crum
Chris Crum has been a part of the WebProNews team and the iEntry Network of B2B Publications since 2003. Follow Chris on Twitter, on StumbleUpon, on Pinterest and/or on Google: +Chris Crum.