What do we know about Google's RankBrain?

Ryuzaki

お前はもう死んでいる
Moderator
BuSo Pro
Digital Strategist
Joined
Sep 3, 2014
Messages
6,244
Likes
13,130
Degree
9
QGJJvja.png


There's been talk about RankBrain for a while, at least since late 2015, but I haven't had the opportunity to dig into it and see what's what. So I wanted to ask you guys, what do we know about RankBrain so far?

My understanding is that it's some kind of side algorithm being informed by machine learning system attempting to better understand the queries being typed in the search bar.

Seems like it's a natural growth from Hummingbird in an attempt to get a hold of semantics and sentence structure. LSI and synonyms are going to become more important and over-optimization thresholds will drop.

If this is only applied to queries (which are already pretty amazing), then it's only a matter of time before it starts being applied to content as well.

The question becomes what are the consequences of RankBrain?
  • Spun Content is going to take a hit
  • Blackhats will start doing content mash-ups...
  • Which will lead to a hit against sites mainly doing syndication
  • LSI and synonyms are going to become more important...
  • And over-optimization thresholds will drop
  • SERP interaction metrics will be taken into higher consideration, like CTR and Pogo-Sticking
What's interesting about machine learning is the iterations of improvement constantly happening. At a certain point, it might jump into the realm of true artificial intelligence. Eventually when it's deemed good enough, I can see this leading to major changes in the core algorithm as well.

Seems like this is an attempt at automating all of the man-labor they were doing to figure out what quality signals mattered. They probably got enough data to aim RankBrain in the right direction and now are going to let the searcher's signals drive it home the rest of the way.

Every solution introduces more problems. Paying microworkers to search your brand and click your links might become a more lucrative activity, although one could argue that it's easier to just write good content with good meta titles and descriptions. This also makes you think that the currently established ranking sites will get a huge lead on everyone else, so there might be more shuffling and testing of the SERPs like the honeymoon period does to get some signals in on each URL.

What do you guys know about this? What are you thinking about it?
 
Reading your article I just see another reminder to put in a real genuine effort to create websites that have useful, quality content and services delivered in the best way possible for the consumer.
 
"Paying microworkers to search your brand and click your links might become a more lucrative activity,..." something like this already launched. Alex Becker created a new software which pushes your site to the top with visitors who search (for a specific keyword) & visit your site. Or something like that. He promises massive results... (I'm not in any way advertising for him)

"although one could argue that it's easier to just write good content with good meta titles and descriptions."
I wrote 2 blogposts this week which totally crush the competition content-wise. One of them instantly ranked on position 46 on a fairly weak website. If I build some links to it, I'm sure that it'll get to the top position. So, I'm totally with you on this one.
 
I think RankBrain, at least at the beginning, will be more Google propaganda than anything else. In the future, maybe spun content will take a hit, but in the short-term there will be no major changes.
 
I'm just wondering when this will backfire and start penalizing legit sites, or even if it will?
 
"Paying microworkers to search your brand and click your links might become a more lucrative activity,..."

I remember reading a few years back that people were doing this in a variety of sneaky ways. IIRC some/most were using Mechanical Turk until they banned those job types. I recall wondering at the time why MT would care either way if the jobs were actually to boost SERP rankings..
 
I'm just wondering when this will backfire and start penalizing legit sites, or even if it will?

My guess is, if the scenario is correct, when they start hammering down on spammers using content mashups as a replacement for spinning (or mashups with slight spinning to keep grammar scores high), there's going to be a lot of original sources taking hits. And there'll be a lot of instances where spammers will yank content out of RSS feeds before Google indexes it and gets it indexed themselves, thus becoming the original (and likely infinitely less trusted though... that might solve itself).
 
Maybe it's just me but isn't everything google has done and is doing trying to emulate a human, well superhuman/freak or somethng like a group of human experts ?, at least that is how I've alsways looked at gooogle and their changes.
 
I think most of us have known for a few years that spun content will ultimately be found out by Google. Unless it's really, really good quality stuff, people can tell it's been spun, and if people can then the google brain will do so soon too. I would never go near spun content on a long term money site now.
 
Maybe it's just me but isn't everything google has done and is doing trying to emulate a human, well superhuman/freak or somethng like a group of human experts ?, at least that is how I've alsways looked at gooogle and their changes.
In a sense, yes. What Google is implementing is Machine Learning. It's also called Computational Cognitive Neuroscience. In Machine Learning, artificial neural networks are made, feed data, and taught what's right and wrong for a task. The network then learns the data and is able to perform the task. It's modeled off of actual neuron behaviors. It's not only Google using Machine Learning. About every other company is too.
 
Back