Given the amount of apparent confusion around how it will affect your marketing efforts, we felt it was time to dive into what it means to you.
Connecting the dots
What I do think is important is that we’re seeing a definitive moment in where search is probably going in the future. Give we have a Google Hangout on this topic today, I felt one more post was in order.
Again, I get the sense that these are early days for the associated elements that seem to be part of this current chain in the evolution of how Google works. We know that they are using the machine learning parts in other areas such as spam detection already. So that part of this puzzle, isn’t limited to just RankBrain.
We also know that the holy grail is voice search (known as conversational search within Google). I can’t help but think that RankBrain, in essence at least, is part of moving towards that direction I’ve added a video at the end for more on conversational search.
If you’d like to learn more about the shift, here are some elements worth considering;
For me it’s important to consider that what RankBrain is today, is likely not where it will end up. The starting point of affecting ambiguous queries, is seemingly somewhere Google can implement the technology and move from there into other areas of search.
It will certainly be far more ubiquitous down the road in voice search, and in a larger implementations within standard text search.
How does RankBrain affect Rankings?
That’s the question that seems to be on a lot of folks minds in the SEO world. From what we do know so far, it doesn’t directly affect rankings. That is to say that it isn’t a set of scoring mechanisms itself.
Google has stated that,
“RankBrain is one of the ‘hundreds’ of signals that go into an algorithm that determines what results appear on a Google search page and where they are ranked,”.
That’s the interesting part. How it’s affecting rankings is that it essentially re-writes the query in a sense, which in turn is going to change what is served back in the results.
As such, in that sense, it is certainly affecting rankings, without being a direct scoring element (such as PageRank). It is probably something I’d consider more of a boosting/dampening signal. Most certainly, at this point, RankBrain isn’t evaluating things on-page. Not so say that this won’t change and we could see a direct ranking factor at some point. For the moment, this is more about query classification.
How will it affect your search marketing efforts?
And so, back to the original question.
When asked on a recent podcast interview; “.. how should SEOs explain RankBrain to their boss or client?” – I answered, “Don’t?”.
That means you (as a marketer) probably don’t need to try and “optimize” for RankBrain. This is more about improving search quality for Google.
This won’t really have a direct affect on how we do things. It’s not really something we can optimize for per se. Which is what someone asking that question, really wants to know. I would suggest re-directing them into a different thought pattern;
Contextually solid content – ensure that the content you’re producing is solidified around concepts, entities and the meanings on the page. Get past the tactic of stuffing desired terms on a page and move towards concepts, phrases, entities and their relations. This is important for far more than RankBrain.
Don’t discount conversational search – while it’s not currently an overly important part of SEO, we want to understand that one of the most important areas in the near future is going to be voice search. You are going to want to ensure you understand it well and account for where we’re headed.
Understanding how RankBrain fits into the evolution of search is the real take-away here. It’s not about how we can adapt to it nor how it’s going to affect rankings and traffic (in the short term).
What is RankBrain?
There’s already been a ton of coverage on this, some of it way off-base and/or click-baity, and some other posts are far closer to truly understanding where this is headed. I think Danny’s original post (last fall) is certainly one of the better ones to use as a starting point.
Ultimately, RankBrain can help Google better understand words and how they relate to each other. Some of the core elements take semantic analysis to a new level by turning words into vectors, instead of the more traditional word and phrase relation approaches.
Here’s at some simple word vector examples;
King – man + woman = Queen
King relates to prince
Queen relates to princess
Given that King relates to Queen,
King can also relate to princess.
The system is looking for word relations and (within a vector graph) how closely related they actually are. By understanding how these words relate, they can better understand what a given query might be looking for.
Here’s an output from a word vector neural net visualization tool;
The red neural paths being the strongest relation, to the blue ones, which are the weakest.
The reason that they’re doing this, instead of traditional semantic analysis, is that the vector approach lends itself to machine learning and ultimately, an artificial intelligence approach.
Now, from what we know, it’s only primarily affecting some 15% of queries (but being run across the entire system). And those queries, are the more ambiguous and long tail ones that Google had problems truly understanding in the past.
Some videos worth watching;
The Future of search (SMX 2016);
Deep Learning; Word2Vec
Word Embedding Explained and Visualized