Machine Learning Affectking Featured Snippets | Rank Ranger

Two words can either make or break your day… Featured Snippets. For those that score them with consistency the SERP feature is a godsend, for those that don’t… well that’s another story entirely. In either case, combining the position zero boxes with yet another two-word term… machine learning… might send some of us into a tizzy. Yet, tizzies aside, that’s precisely what I intend to do as I strongly believe that Google’s machine learning properties are touching Featured Snippets in all-new ways.     

Where We Stand With Featured Snippets 

Before we get into the thick of it all, let’s first get a greater understanding of where we find ourselves with Featured Snippets. On
desktop, in the US, the feature is found on about 12% of all SERPs, which is considerably more than on mobile where we’re seeing Featured Snippets on about 7% of search engine results pages. 

The SERP Features Tool showing the percentage of page one SERPs that contain a Featured Snippet 

As you may have suspected, Google prefers to pull Featured Snippets from the top organic result on the page and if fact does so about 34% of the time. Perhaps unexpectedly, Google pulls content for its position zero boxes from the 4th position 13% of the time, and the 5th position a surely surprising 8% of the time. 

The top two positions on the SERP are responsible for the content within more than 50% of Featured Snippets

Of course, like any average, there is data that is closer to either end of the spectrum. Here, the Medical, Pharmaceuticals, and Biotech niche see its Featured Snippets pulled from the top position about 44% of the time (10 points above the overall average):

Conversely, the Travel industry has its Featured Snippets pulled from the first ranking position but 25% of the time (nearly ten points below the average):

From a purely quantitative point of view, not much seems out of place when it comes to the most awesome of Google’s special search features. However, a closer look at Google’s recent behavior vis-a-vis Featured Snippets tells an entirely different story. One that shows significant shifts in how Google is able to satisfy user intent, and even sub-intents via the all-powerful top box. 

==> Find out how to get started with machine learning in SEO

Featured Snippets and Machine Learning – What We Know & Where It Leads Us 

Asserting that machine learning plays a significant role in the world of Featured Snippets is old news. Which is why I’m not making any such declaration. Instead, I’m claiming that the way Google utilizes machine learning when serving up its Featured Snippet content has advanced… dramatically. Thus, and as good prudence would demand, let us refresh ourselves and look at the role of machine learning vis a vis Featured Snippets, as we currently know it.

A Machine Learning Sentence-Compression Algorithm for Featured Snippets

Ever since way back when, which in the world of SEO is two years and beyond (in this case 2016 in specific), Google has used machine learning to create the content that appears within its Featured Snippets.

In simple terms, Google’s machine learning properties work their way through a piece of content and string together a set of sentences in order to serve you a relevant answer within any given Featured Snippet.

There are two things that are important to note: 

1) Google’s machine learning properties are an integral part of analyzing the content found in Featured Snippets.

2) The above information is old news from a time when Google just started allowing machine learning to heavily touch its “search properties.” 

I can understand that the former is significant (though not groundbreaking news), but the latter, that machine learning has impacted Featured Snippets since 2016? Who cares? Why does it matter that the news regarding machine learning sentence-compression algorithms generating Featured Snippet content is old? 

Simple, because machine learning, by its very definition, is meant to evolve. That is, the entire point of using AI, and in this
case the particular strand of it known as machine
learning, is that it evolves and becomes “smarter.” Thus, if the news of Google’s machine learning content generation inside of Featured Snippets broke yesterday, there would not be much to add. However, it being the case that this is “old news” from 2016, the question begs itself… how has machine learning evolved in relation to Featured Snippet content? 

The Evolution of Machine Learning’s Impact on Featured Snippets 

To get started, let’s take a step back for the sake of gaining some perspective. One of Google’s main tasks it has given machine learning, if not the main task, is to better understand user intent. Now, when I mean “user” intent, I not only mean deciphering the intent of a search query. That’s only one half of the equation. To adequately match the intent embedded within a given search term, Google also has to be competent at understanding both the latent and manifest intents within a given piece of content. Otherwise, the search engine would be relatively incapable of matching the intent of the query with highly relevant content. 

It stands to reason, that over time, and since the inception of Google AI properties such as RankBrain, that the search engine is better able to understand
intent. If you were then to combine Google’s increased ability to better decipher intent, with an evolved ability to construct content within the Featured Snippet, you might get something like this: 

What you have here is the content of a Featured Snippet combined with the answer style used in Google’s Direct Answers. Google, understanding quite well, it would seem, that the entire intent here is to see a dollar value, pulled out the cost from within the snippet’s content and prominently placed it front and center. 

This Featured Snippet format is a relatively recent incarnation that takes the bolded content traditionally found within the position zero box a step further. That is, Google has long tried to accentuate the content within a Featured Snippet that it deems highly relevant. To do so, the search engine simply bolds the words or phrases it has determined to be the most pertinent:

You might argue that the
above Featured Snippet offers multiple dollar amounts, and as such Google cannot offer one figure to place above the content itself. It’s a good argument. That is until you get
to a Featured Snippet such as:

As you can clearly see, there is no range of pricing and there is no header as in the case of the snippet for
walk in tubs cost.

It’s clear that Google has increased not only its ability to interpret
intent, but to pull content from a larger piece of text that satisfies such intent in a hyper-specific manner. That is, it is not a simple trigger (i.e., when no dollar range is listed the price is displayed as a header). Now, why did one snippet get the header and not the other? Good question.  

Google’s Obvious Turn Towards Machine Learning for Its Featured Snippets 

The truth is, I didn’t have to go this roundabout way of showing you that machine learning is a much bigger part of the Featured Snippet equation than it once was. If I were not intent on offering you a comprehensive look at the machine learning situation as it relates to Featured Snippets, and if I did not want you to build your understanding progressively, I would have just told you about Multi-Faceted Featured Snippets:

In the example above
Google both meets a specific intent by offering a Featured Snippet discussing the math rule per se, as well as a more general intent by showing a list of the “properties of addition and multiplication.”

It’s relatively easy to see how Google, via its increased intent interpretation capabilities, has let machine learning loose on Featured Snippets here. However, in case that wasn’t obvious enough, Multi-Faceted Featured Snippets are not coincidentally similar to Bing’s Multi-Perspective Answers which, as opposed to offering one true answer, presents multiple snippets with varied information. When it comes to the Bing property, there is no doubt that AI is at the steering wheel, Microsoft has proudly said as much (which makes you wonder why Google has all but admitted that machine learning drives Multi-Faceted Featured Snippets, yet has not officially done so).

Machine Learning, Moving Beyond Multi-Faceted Featured Snippets 

The concepts that underly the Multi-Faceted Featured Snippets don’t exist in isolation. The whole idea of the Multi-Faceted Featured Snippet is to give you access to content that may better align with your actual intent, or meet a sub-intent that exists within your overall aim and purpose. The underpinning of doing so is Google’s ability to comprehend how various topics or sub-topics all relate to each other. It’s a concept, or rather an ability, that is at the very foundation of Google’s surging forward in its quest to meet intent via its Featured Snippets.  

How do I know that the basic concept that is at the root of Multi-Faceted Featured Snippets is pervasive? Take a look at these three Featured Snippets and see what you notice about them: 

Keyword: Find Gold

Keyword: How to Find Gold 

Keyword: Where to Find Gold 

Same title…. but
different content within the snippets per se. 

The interconnectivity of the content shown in Featured Snippets (as manifested by a uniform title for multiple pieces of content) points to the interconnectivity of the neural networks that comprises machine learning. Google clearly understands that there are multiple sub-intents that fall within a larger context. If I could offer a general rule of sorts, when you see Google dissect intent by sub-categories, you can bet your bottom dollar that machine learning is at play. We’ve seen this with organic rankings and now we’re seeing it with Featured Snippets. 

Google Goes Deep With Intent Inside of Featured Snippets 

To sort of cap off this little journey into how machine learning is making a serious mark on Featured Snippets, let’s talk filters. Google has a long tradition of experimenting with various kinds of filters with the aim of pointing you towards related topics within the Featured Snippet. As of April 2018, it would appear that Google has decided on a filter that applies to an extensive set of Featured Snippets.  

Great, so Google can show you a set of related topics within the Featured Snippet, why is that so significant and an important part of our conversation here? 

Because the filter within the Featured Snippet is not just a set of related topics for you to peruse through. 

Have a look at the filters that show up for the keyword bathtub cost: 

Did Google offer me a list of other topics such as bathroom sink cost to explore? No. What it gave me was a way to refine my intent. Did I mean to look at the cost of a jacuzzi? Was I looking to buy a new tub or simply refinish the one I currently have? The filters are not multiple topics, they are intent refinement mechanisms.  

In this way, the newly released filters are very much similar to the machine learning driven Multi-Faceted Featured Snippets. They are attempts by Google to better meet your intent and are the result of machine learning’s ability to better gauge your true query desires (that sounds so ******** doesn’t it?). 

Take a look at another example, this time for car maintenance cost: 

Did I get suggestions within the filters for motorcycle maintenance cost or most expensive car to maintain? No, because this filter is not topical, it doesn’t work on a horizontal axis but instead aims to drill deeper. Here, Google picked up on the fact that my query was most likely too broad for what I had really set out to achieve. In other words, why would I execute such a search? It’s either that I don’t own a car and am looking to buy one and am wondering about the associated costs, or perhaps I’m setting up a budget and want to know how much to allocate… anything along these lines. As a result, I most likely need to know the cost of car repairs within a given time frame, a month, a year, etc. 

Why Sites Should Be Concerned by Machine Learning’s Featured Snippet Presence… Very Concerned 

Great, your site has its URL showing
within the Featured Snippet at the top of the page. But what if that’s not you, and your URL is not located within a Featured Snippet? Even when it is, what are the chances you score all of the applicable Featured Snippets across all of your targeted keywords? What then are the implications of machine learning’s proliferation in regards to Featured Snippets? 

It means that Google is far more aware of what the user wants. Whether it’s Multi-Faceted Featured Snippets or a prominently displayed dollar amount for the expected cost of a walk-in bathtub, Google is showing highly relevant and highly specific Featured Snippet content. But isn’t that a good thing? For users yes, not for the fourth result on the SERP offering financial tips when undertaking to install that ever-popular walk-in tub.

This becomes more obvious when looking at the current “intent filters” used in many of the Featured Snippets you’re likely to come across on the SERP. In this instance, users, when trying to refine their own query intent, can hop from one Featured Snippet to the next until they find the
postion zero content (and the associated URL) that fits their exact want and need. That is, instead of looking down the SERP for sites that are more relevant than the Featured Snippet may be, users, are moving on to the next top-ranking box on a brand new SERP. In such a scenario, scoring Featured Snippets for a series of keywords is extremely important. However, that’s a very difficult task. 

This however only skims the surface in understanding the real impact of machine learning having more influence over Featured Snippets. That is, filters aside, Google is becoming quite adept at showing Featured Snippets in a way that has a greater chance of satisfying what users set out to do, and it owes this success to machine learning. Consider a query for snake bite types:

Google’s neural network is very much aware that the user in this instance is not referring to a generic topic, but one that changes from one species of snake to the next. More than that, Google knows that a user is highly likely to want to see an image, as having one is extremely relevant when dealing with the biting serpent, as it’s the only way to really identify one type of snake from the next. In other words, Google determined that the query’s intent entailed multiple sub-topics, in this case, multiple snake species. It then further realized that in such
a case images were necessary. As a result, Google offers a carousel of snake images according to species within the Featured Snippet. Engaging with any of these images brings up a new SERP associated with the snake that is led by a Knowledge Panel, i.e, a Google property. 

In knowing, to a far greater degree, what the periphery of the intent picture looks like, Google can use a powerful combination of its SERP features to cater to the user’s “true” wants. The better Google gets at this, and the more it uses multiple SERP features (as it does in this case) to do so (which I plan on discussing in the near future), the harder it will be for the other ranking sites to see a glimpse of daylight, let alone have their day in the sun.

Are Featured Snippets Still a Win? 

Google’s machine learning properties seem to have turned another corner, and as a result, are touching the Featured Snippet in new, and highly specified ways. As this trend continues, it only will be harder for URLs that don’t appear in the position zero box to attract site traffic. Google’s machine learning properties have the ability to make any given Featured Snippet that much more potent.  

Let’s consider the other side of this for a second. Featured Snippets are a win for a site. With that, let’s remember the real goal, to have users click on your URL. If Google is getting better at meeting intent within the Featured Snippet per se, does that bode well for your site’s traffic? 

Now throw in carousels that lead to Knowledge Panels or filters that lead to other exceedingly relevant and potent Featured Snippets. How do these elements factor into your site’s traffic via a position zero box? 

Featured Snippets used to be a nice tradeoff. Google would pull a chunk of your content to display on the SERP, making its users happy, and you would get a nice flow of traffic to your site, making you happy. Does this exchange still hold up as machine learning continues to increasingly influence Featured Snippets? For now, yes, but going forward is another thing entirely. I think we’re heading into some tricky snippet waters and we’re all out of life vests. 

Check out our guide to Featured Snippets to learn more about the different types of Featured Snippets and how to obtain them.

About The Author

Mordy is the official liaison to the SEO community for Wix. Despite his numerous and far-reaching duties, Mordy still considers himself an SEO educator first and foremost. That’s why you’ll find him regularly releasing all sorts of original SEO research and analysis!