Sign up for Mazen early access & updates

The first SEO software that saves your time
SEO guide
Chapter 4

Semantic On-page Optimization: Optimize your content for RankBrain and reach the first page of Google with half the domain authority of your competitors

In September 2016, Google filed a patent for context vectors that would improve search. In this article, I’m going to (briefly) explain what this patent can tell us about RankBrain (Google’s AI algorithm) and how you can easily gain a huge SEO advantage over your competitors.

The website SEO by the Sea published a very interesting article about the patent and cited Bloomberg’s in-depth intro to RankBrain titled Google Turning Its Lucrative Web Search Over to AI Machines:

RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand.

In layman’s terms, this basically means that Google found a way to convert any word or combination of words into vectors, which appears as a number sequence in the source code like: 0.50882536.

word2vec SEO simple explanation

How Google converts words into vectors. For a more scientific explanation, you can check this presentation.

The goal of this RankBrain update was to be able to compare the lexical proximity between a cluster of terms. For example, “camel,” “hot,” and “desert” have a pretty close lexical proximity. On the other hand, “cheeseburger” and “semantic” have a very low lexical proximity.

The closer the “number sequences” representing each of the words analyzed by RankBrain are to one another, the more likely those words are “talking about the same thing”.

What’s more, Google can use this vectorial analysis to identify words that are only related in a specific context.

For example, to a rancher, a horse is an animal. To a carpenter, a horse is a work surface. To a gymnast, a horse is an exercise apparatus. Vectors help the search engine identify pages that are semantically relevant to the user’s original request.

In the patent itself, Google describes it like this:

This system may be used to find information or classify information by subsequent inputs of text, in calculation of macro-contexts, with ultimate determination of lists of micro-contests including terms closely aligned with the subject matter.

 

Makes sense, right? Right...? In English, what that means is that, in order to find information, Google is going to look for terms closely related to the initial request…

We call this the “semantic relationship.”

And since we care about your success, we’ve put together a step-by-step guide to optimizing your content for Google’s RankBrain.

Let’s dive right in.

Explore how Google RankBrain interprets the meaning of your content

If I give you this sentence:

I'm a beautiful and proud beast with four legs. I can be brown, black, white, or many other colors. I was one of mankind’s first friends and, for thousands for years, humanity has used me to move from one place to another.

It should be pretty obvious to you that I’m talking about a horse, right?

So let’s see if Google understands it the same way. Copy and paste that sentence into Google and then click on “Images.” You results page should look something like this:

Clearly, Google did not understand that we were talking about a horse. Why might that be? 
RankBrain Google Image Search

Simple. There isn’t a single word in that sentence that is semantically connected to the word “horse.” Our super-duper human brains, on the other hand, can make that connection because of our deep cultural understanding of the role horses have played throughout human history. Plus, horses are a part of our basic cultural knowledge.

Now, let’s try something a little different with our dumb friend RankBrain:

I'm a beautiful and proud animal. I eat grass and live in a meadow. I was one of mankind’s first friends and, for thousands of years, humanity has ridden me to move from one place to another.

Here is the result:

 As you can see, the words in blue have made a bit of a difference
RankBrain Google Image Search 2

Google now better understands the context/intent of the request and the results a more pertinent. However, there are still some signs that the search engine doesn’t fully grasp which animal we’re talking about. Google needs some more precise “semantically related” words.

This time, we’re going to use terms that are more related and specific to the word “horse”:

I'm a beautiful and proud animal. I can be chestnut, black, white, or many other colors. I eat grass and gallop in meadows. Humans ride me using saddles.

Now, finally, Google gets it right:

RankBrain Google Image Search 3

So what does this experiment actually tell us? Well, there are three very important takeaways:

  1. Writing great content for RankBrain means using specific words that help the algorithm identify the intent and purpose of the page
  2. Just because people understand your content doesn’t mean Google will.
  3. If you want to rank better, create a semantic field that shows Google your page content is accurate, precise, and trustworthy regarding the topic at hand (i.e. your main keyword). The more specific and sophisticated your semantic field is, the better you’ll rank for the chosen keyword.

Let’s take another example.

On the very competitive request “cheap kitchen tables,” the page https://www.americanfreight.us/dining-rooms is ranked 7th on Google in the US (July 2017). However, that page has a lower domain and page authority than the ones that rank 8th to 15th.

Results 6 to 10:

Domain authority and page rank influence on ranking example

Results 11 to 15:

Domain authority and page rank influence on ranking example 2

Here is what the page looks like:

American mattress SEO example RankBrain

Now, you won’t see this page win any design competitions, but take a look at the bottom of the page; there’s a short paragraph clearly dedicated to SEO. And you know what? It does its job.

The paragraph brings great semantic value to the page and includes lots of words from the semantic field of “cheap,” “kitchen,” and “table.” Let’s explore this semantic field in more detail.

Semantic SEO writing copy example

Let’s be honest, this paragraph sucks. Who could possibly want to read that? Fortunately, it’s not meant for you and me to read, but for robots instead. I personally think this low-quality text borders on keyword-stuffing, and should probably be penalized for it, but it seems to have remained within the limits of what Google deems acceptable. The writers obviously employed semantic variation with the keyword “cheap”; there are five different words from that semantic field (cheap, discount, lastest deal, lowest price, as low as).

Okay, so that last example was a little extreme.

I’d now like to show you some content that I think strikes a better balance between being useful to your audience and relevant for RankBrain.

For the very competitive request “ultraportable laptop,” this page (website http://www.ultrabookreview.com) is ranked 6th in Google.

As you can see below, the pages that rank below this result all have very high domain authority (over 70/100) and higher page authority than the ultrabookreview.com page.

Semantic field and ranking example

But if you take closer a look at the page, it’s obvious why it ranks so well: it contains 13,000 words of ultraportable laptop reviews. These reviews are so extensive and complete that the semantic field for “ultraportable laptop” is simply much richer than on rival pages. That’s why this page can rank on a very competitive keyword despite having a weaker link profile than the competition. Not to mention all the long tail phrases and many variations of the main keyword (“ultraportable laptop”).

How to optimize your content using semantic fields

Ok, so now that you understand the importance of using related words when crafting your content, how the heck do you actually choose the words to put into your semantic field? I’m glad you asked:

  1. Type your main keyword for the page you are currently optimizing.
  2. Visit each of the first 10 pages, specifically the ones that have lower domain or page authority. You can use MozBar to display backlinks and authority info directly in the SERP.
Horseback riding SEO example optimize semantic

In this example (keyword “riding summer camp”), I visited the following pages:

https://www.forrestel.com/

http://www.horseridingcamp.com/

http://www.petitgalop.com/Summerridingcamp.html

https://rockbrookcamp.com/horseback/

  1. Next, examine the text on each of these pages and note the words that you think are clearly semantically related to your main keyword (like we did before with kitchen tables). I’ve done this exercise for the pages above and split the semantic field into five different categories to better illustrate the process:
Semantic SEO vocabulary example

The colors indicate the page from which I took each word. Each column contains semantically-related words in the content. If you were to only read those words, with no images or other context, I’m pretty sure you would be able to tell exactly and precisely what each page was about. In a nutshell, that’s exactly what a search engine does: it identifies meaningful words so that it can understand the content and recognize similar content.

  1. Finally, craft your content using the words you collected. This will help you precisely describe the ideas, objects, services, places, and events that you are talking about.

 

Are you tired of hearing the same advice about writing engaging content? Are you having a hard time getting backlinks in your industry? This semantic approach could change your SEO life. So, please, give it a spin and share your results here. I’d be really interested in getting your feedback.

20844ba0ad8c0f071b64e5c5c13e48fbffffffffffffffffff