In our main article, here, we discussed how RankBrain went from being a pilot project to becoming one of the primary ranking factors in Google’s algorithm. Now let’s talk about the implications for SEO of this takeover by an artificial intelligence of humankind’s most-used search tool.
Whenever a major algorithm update hits Google, we usually ask two questions:
How will this make things different?
What can we do to optimize under these new conditions?
With the advent of RankBrain, there are a number of experts who believe the answer to the second question is ¨nothing.¨ They either believe that RankBrain is operating on the periphery and that SEO should stay focused on on-page factors and off-page factors that are well-understood, or they believe that it is there is literally nothing you can do to make things better on RankBrain so you should just focus on content.
Of course we disagree with both of those positions, but we will get back to that after we look at the first question.
Nobody knows exactly what RankBrain is doing, but we think it is trying to understand web content as a human would. In theory this means that RankBrain will continue to be more and more human-like and SEO will become more and more an extension of user-experience and general marketing. But some changes can be predicted still:
One of RankBrain’s superpowers is to match new phrases to existing phrases. But why limit this to new phrases? If it is true that RankBrain now looks at every search, then we can assume it is matching all phrases to each other. So highly ranked content for one particular phrase now becomes highly ranked for all related phrases. This sounds like a good thing, you can create content that will rank very well for keywords that it does not include. But a lot of SEO is built on finding less competitive versions of phrases. So if the top ten slots are filled when someone searches for ¨divorce attorney,” maybe you can find room at the top for ¨family law practitioner.¨ But if RankBrain is busily matching those two queries together, then everyone is competing for the same ten spots again. Long tail SEO may not be completely dead, but there is a lot less room for it now.
Another feature of machine learning is that it is constantly changing. The past relationship with Google has been punctuated with large algorithm changes, usually with cute names like Penguin or Panda. The cycle went like this: SEO pros worked out as many details of the algorithm as possible, sometimes helped by Google announcements. These details were put into practice as changes to websites and content. Meanwhile, the blackhat and grayhat specialists would find ways to exploit these details in ways that go against the reasons for those changes. Eventually spammy links and results would be enough of a problem that Google would release a new update. Websites built entirely on exploits would crash in the rankings, while those that sacrificed too much quality to match ranking factors suffer. Then the cycle starts again as everyone tries to figure out what changed. RankBrain changes will not be like this. Machine learning is more gradual, iterative. Coincidentally, Google has stopped announcing algorithm updates more or less since RankBrain started to take over. SEO will be less like NFL style football and more like Soccer or Basketball.
An open question is the relationship between RankBrain and Leapfrog, Google’s human rating project. There has been no public link made between these two projects, but why would RankBrain not use this huge body of valuable data on human responses to search results? If so, we could see the algorithm be even more responsive to human behavior. On the other hand, it is possible that human behavior isn’t as unique as we think. Leapfrog is already known for pushing the participants towards a very particular way of thinking about results, and recent news is that it is being scaled back now.
It is almost certainly wrong to say that RankBrain will end the need to focus on meta data, titles, or anything other than that content itself. The central problem with search has not changed, there are billions of documents out there and we need to sort through them all and just find a few. Content structure is not just a way to outsmart search engines, it allows content to classified and sorted in the first place. We used structured data to help manage this flood long before search engines were invented. RankBrain will not eliminate the value of having a title for an article.
So What Can We Do About It?
The obvious thing is to remember that RankBrain is the third most important factor. The first two, content and links, are still critical. And content structure is very much still part of optimization.
Another place to focus is on the snippets of your content that may appear in search results. This is normally either your meta-description tag or through structured data using recognized schema. The key here is that RankBrain is probably sensitive to click-through rates, and the information you present through your snippets acts like a call to action.
Keywords, individually, may be less important, but focusing on clusters of keywords or topics can increase your reach. In the past, many SEOs have preached that content should be chunked. But as RankBrain grows in importance, we can talk more about central pieces of well-organized content, as long as the structure is inviting to human users, of course.
Pay even more attention to user experience. RankBrain is trying to simulate the experience of human visitors and will probably be tracking any available indicators of user satisfaction. How it does this will not be clear, because it is basically creating its own algorithm. There may even be human raters involved in some fashion. Optimize for the user and you will be optimizing for RankBrain.
Finally, pay attention to your analytics. In the past, SEOs have liked for abrupt changes timed to recent major updates, but the constant slow change of machine learning will require closer attention to trends.