Google has been working on artificial intelligence (AI) for several years. When I first reported Google attempts at AI back in February 2012, it was still a far flung dream. Since then Google has achieved an early stage machine learning artificial intelligence that has improved search results, and created higher than expected profits for the company this year.
Normally, significant changes to Google's search ranking methods are noticed by those of us in the search engine optimization industry, but this time it turned out that Bloomberg News got the scoop on all of us. A few weeks ago, Google hit record quarterly earnings and per share prices. Naturally, Bloomberg asked how they did it, and the answer surprised us all.
The answer was to place the trust part of its business in the hands of AI.
Creating Search Results
The results you see in Google search are generated by a complicated set of equations known as the Google Algorithm. Over the years, the Google engineers have programmed individual ways of analyzing websites based on mathematical models. Search results typically look bad when the question you answer can't neatly fit into one of the pre-programmed equations. This is where RankBrain will be triggered.
RankBrain is designed to understand colloquial language and phrases that have never been seen before. It will look at similar phrasing and make a guess at providing answers rather than just relying on the mathematical equations.
Google normally deploys their new search algorithms slowly, and those of us who pay close attention to ranking results, notice when things have changed. But in this case, Google has already deployed RankBrain throughout the world and it wasn't until they announced it last week that we even knew about it.
What is a Google Search Signal?
Originally, Google thought that RankBrain would help a small fraction of the daily search results, but according to this interview on Bloomberg News last week, it's the 3rd most important ranking signal for search results today.
Google representatives are always claiming that there are hundreds of different signals that make up their ranking algorithm. Each signal is one of those pre-programmed equations that engineers created based on years of monitoring users. We don't know which of these signals help a website rank higher than another, and one signal will may play a more important role at ranking a site today than yesterday. The QDF signal, for example, only activates when an unusually high number of requests flood Google search in a short period of time. When that happens, the QDF signal will take precedence over most of the others.
I just spent a few hours reading, watching, rereading, and re-watching all the information about RankBrain since it was mentioned on Bloomberg last week. I find it a little confusing that it is being called a "signal" in the normal sense, when it seems like an alternative path for determining search results.
When Was Judgment Day?
The Bloomberg interview also revealed that RankBrain was fully deployed a few months ago after more than a year of planning and a green light in early 2015 was given to for inevitable rollout. Looking back over my calendar, I realize now that the final rollout probably coincided with the massive Google My Business, Google Maps, and Google Local 3-Pack revamp that took place in early July 2015. While the SEO world was complaining about those changes, none of us noticed the affect RankBrain was having on normal search.
I can't help but wonder if Google intentionally planed the deployment of AI during those other sweeping changes just to smokescreen the more important changes to search. Just stop and think for a moment about this... While Google doesn't use scary names like Skynet, they have, in fact, secretly deployed an early stage AI that's being used worldwide.
Don't get me wrong, I personally feel that all the sci-fi horrors portrayed in movies has helped to build awareness and guides engineers worldwide towards creating a Star Trek AI computer rather than a Terminator AI computer that summarily passes judgment on us all. I do hope, however, that this AI will help identify duplicate content more accurately.
RankBrain and Hummingbird as Parallels
Google rewrote it's algorithm from the ground up back in 2013 and launched it as Google Hummingbird in August 2013. Search results have improved greatly since then, but Google is forever changing and tweaking their equations based on user actions they measure every day. But all that tweaking requires hundreds of hours for Google engineers to analyze and test their findings before deploying changes.
On the other hand, RankBrain is a machine learning system that continually learns. Another reason I'm not viewing this as a "signal," but rather as an alternative path to generating search results, is because this post on Bloomberg mentions a contest between RankBran and Google search engineers to see who could guess how pages would be ranked. RankBrain had an 80% success rate at guessing while the engineers only had a 70% success rate. For me, this is a clue that RankBrain is a self contained, parallel algorithm.
How Does This Effect You Today?
The Google engineer interviewed on Bloomberg specifically mentioned that RankBrain was designed to help understand colloquial language and search queries they had not yet seen. I interpret this to be a better matching technique between strange long tail keyword searches to what you have on your website.
It all comes back to what you have on your website. You're always more likely to appear in search results for random long tail searches when you have a lot of information on your site. If you sell a lot of different products and provide several services, then you need a lot of content on your website.
You can succeed with a single web page if you only sell one widget or one service, but for the rest of us, the best practice, even with RankBrain, is to continue adding content to our sites on a regular basis.
