fbpx

AMA with Gary Illyes: Rank Brain is 3rd Largest Ranking Factor

by | Jun 24, 2016 | News | 1 comment

by | Jun 24, 2016 | News | 1 comment

Gary Illyes recently sat down with Danny Sullivan for the yearly SMX Advanced AMA session with the head of Google Search. The session was packed with SMX attendees (me included) hoping for any insider information Gary was willing to expound upon.

Danny Sullivan did a great job keeping the mood light while taking some jabs at Gary to try to get some clarification on issues that the SEO industry as a whole are confused about. Here are some of the things that Gary expounded upon during the session:

RankBrain is Used on Every Single Search

Danny asked Gary if Google was still applying RankBrain to 15% of its searches. Gary explained that RankBrain is used on every single search query now.

This is a huge jump from the previously reported 15% that Google gave us a few months ago, but Gary explained that Google is committed to the machine learning algorithm to better serve up more relevant queries.

How Does RankBrain Work?

While these questions clearly frustrated Gary, he eventually did give some real life examples of how RankBrain is working behind the scenes.

To better illustrate the basics, Gary held up his water bottle that he was drinking from and asked the audience what it was. Shouts of ‘plastic water bottle’, ‘bottled water’ and ‘water’ rang out. He then asked how we came to that conclusion. By looking at the water bottle, recognizing the label and seeing that the liquid inside of it was clear, is how one comes to the assumption that it is bottled water.

Gary explained that RankBrain is learning similar identification patterns and then buckets the data. It then uses this data to seek out past searches and based on what has worked well for those searches, it tries to predict which results are best for a certain word.

He also explained that RankBrain works really well for queries they’ve never seen before – which are typically long tail queries. This means that instead of eliminating stop words from the queries, RankBrain actually includes them.

Here is an example he gave:

“Can I beat Super Mario Bros without the walkthrough?”

Instead of the normal algorithm pulling up walkthroughs for Super Mario Bros., RankBrain would actually serve up results of how to solve the game without using a walkthrough as using the stop words actually gives clarity in this result.

Is Machine Learning the Future of Google?

Gary explained that Google is committed to machine learning and is incorporating it into all of their product offerings.

In the case of search, Hummingbird brought better clarity to the entities in a search query, and RankBrain is designed to fetch better results.

On a side note, he said the movie “Her” was a good movie about machine learning and that it’s a decently realistic representation of it.

Is There a RankBrain Score?

I think what Danny was getting at with this question is does RankBrain give each page a score to use in ranking.

Gary said this:

“There is no RankBrain score. You don’t optimize for it – you can’t. Keyword stuffing content will almost certainly not be good for you as it is bad for RankBrain.”

Will all Manual Interaction Stop with RankBrain?

I think Danny was referring to this new algorithm turning into SkyNet, but Gary had this to say:

“It’s still very important that we can debug queries. We don’t ever want to get to a stage where someone sends us a bad query, and we can’t debug it because it was made by a black box. We are extremely cautious with it and experimenting with it.”

Do TLD’s Make Any Difference?

Danny gave the example of daily.news.politics as a hypothetical URL. His question was if Google drops the “politics” off of the URL and since it has the word “news” in it if it would potentially rank better.

Gary clearly got frustrated, but he said that TLD’s do not play a role in how they calculate relevancy for content. The weird part about his answer was that he then explained that Google does have certain cases in which they look into it, but most cases would not apply. Danny tried to get him to further clarify, in which he didn’t, but he did end his answer with saying that buying keyword rich TLD’s doesn’t matter at all.

Panda is Part of the Core Algorithm

Panda is part of the core Google Search algorithm now. He explained that it is not real time, however, it is continually running and collecting data. Once they process the data, they roll out manual refreshes and then rinse and repeat.

When asked when the next refresh was coming, Gary politely declined to answer.

Social Signals are Not Used in Ranking

Gary explained that they don’t use social signals because social networks have the power to ‘pull the plug’ so to speak. Since they cannot control this factor, Gary claims that Google does not use social ranking signals.

Is Google Plus still a Ranking Signal?

Gary spent some time talking about experiments that Google does and how some of them work and some of them fail. He gave the example of how they have stopped giving power to authorship and said that Google Plus no longer gives any ranking weight like it used to.

We shall await the obituary for Google Plus any day now. 🙂

Is HTTP2 Important to Rankings?

HTTP2 is extremely new and he said not to rush it at all. He did say if you do switch to HTTP2, make sure you have a fallback to HTTP in place or else you will ‘shoot yourself in the foot’ as many browsers still don’t handle it correctly.

He said they are actually working on a Google Bot to accurately handle this.

What Should SEO’s be Working On?

Gary listed two things that the industry should be focusing on:

  1. Google AMP – Gary said that this cannot be ignored. Page speed is incredibly important as median page speed for mobile using AMP is .7 seconds, which is what users expect.
  2. Assistants and Chatbots – He said these are going to be huge and that you will want to be among the first people who implement these types of features.
Greg Secrist
Follow Me