What is BERT? And Why This Update Really Matters

Google’s latest algorithm — Bidirectional Encoder Representations from Transformers, or BERT for short — is designed from the ground floor up to handle natural language queries. Did you know that the AI-enhanced algorithm is estimated to currently handle 1 in 10 of all search queries? This is the most fundamental change to Google’s search platform since the release of RankBrain back in 2015. 

BERT officially began its tour-of-duty the last week of October handling English language queries. Although a global live date hasn’t yet been set, it’s expected to come online progressively. As each country and language meets internal beta-testing metrics, the collection of algorithms and processes known as BERT will grow to encompass more and more queries. Does the name, Bidirectional Encoder Representations from Transformers, leave you scratching your head? Yep, ditto for us. Web designers, bloggers, and content writers don’t speak machine language, yet — but we do need to know how BERT impacts SERPs.

Heads up, website and social media content writers! Here are 5 things you need to know about Google’s BERT update…

How Does BERT Impact SERPs?

In a recent press event, Google used this excellent example to demonstrate a real-world use case. In the query “2019 Brazil traveler to the USA need a visa,” the word “to” and its relationship to the other words are essential in parsing the meaning of the phrase. Prior to BERT’s emergence on the scene, Google would not have been able to compute the importance or weigh the variables within the connection. Unfortunately for the user, the monolithic search provider would have returned a SERP related to American citizens traveling to Brazil.

As a Google representative recently explained, “With BERT, Search is able to grasp this nuance and know that the very common word ‘to’ actually matters a lot here and we can provide a much more relevant result for this query.”

Here’s another query that illustrates the big brains behind BERT: “do estheticians stand a lot at work?” In the past, Google’s standard algorithm would have replaced “a lot” with “alone,” ultimately changing the meaning of the query and yielding vastly different results. With BERT, Google’s algorithm now understands that “a lot” in the phrase isn’t a mistake and is referring to the physical act of standing on the job and returns useful results.

Underneath BERT’s Hood is a Powerful Neural Network

BERT is actually a powerful neural network that processes natural language. A neural network is a series of equations that attempt to understand the complex relationships between variables in a way that is similar to the human brain. Although the technical aspects are literally beyond rocket science, networks of this sort can be described using a couple of talking points.

Neural networks can readily adapt to input criteria that are constantly changing. Here’s a perfect example of why this flexibility is revolutionary and so important. Usually, you call your Labrador retriever in for lunch by blowing a whistle — but for some reason — you just can’t find it today. Instead, you yell, “Daisy, it’s time for lunch. Come and get it!” It turns out that your dog, smart as a whip, knows when it’s lunchtime and so you could call her with the whistle, your voice, clapping your hands, or even just making a lot of noise pouring the kibble. (Truthfully, she’s already drooling at the back door.) A lot of us content creators and social media mavens (AKA pet lovers) work from virtual offices at home, so I know you know what I’m talking about. 

The second aspect of understanding BERT, and neural networks generally, is that the output criteria remains the same despite major changes to the input. It doesn’t matter if we’re using the whistle or our voice to call the dog, the result is going to be the same: your pup comes inside from the backyard, enjoys her lunch, and then takes a nap on the sofa as she does every day. Essentially, BERT (and algorithmic processes like it) help our machine-counterparts understand the human language a little bit more like, well, humans do.

What does this look like in practice, though? BERT helps the Google search query engine better understand the overall context and subtle nuances of words in search strings. The artificial intelligence under the hood is then better equipped to match those queries with results that are — surprise — actually relevant. And from a content writing perspective, this couldn’t be more important.

How Do I Optimize for BERT?

According to Alphabet, content providers and website designers can’t really optimize for BERT. They suggest that we should instead focus our efforts around optimizing for people. So while this may not require a huge change in our SEO methods, there are a few steps we can take to assure that our content is still reaching our intended audience.

In a nutshell, BERT will either continue to return relevant results related to your keywords or it won’t. If your content stops appearing for those long-tail queries, traffic will take a nosedive for those landing pages, and subsequently, your SERP relevance will take a ding. However, BERT is adding to its machine-learning engine every day. If your content is no longer appearing as relevant, guess what? It probably wasn’t quite as optimized as you originally thought. You may see less traffic to those specific landing sites, but conversely, you can expect higher engagement and slightly boosted CTRs.

The takeaway is that Google, assisted by BERT, is finally recognizing the intrinsic value in your posts.

How Does BERT Affect My Clients?

From the preliminary data, BERT is most likely to affect your clients with regard to long-tail queries. By tracking phrase-length queries you should be able to see how well your client’s website is responding to the new algorithm. While tracking keyword performance, be sure to include long-tail queries in your preferred tracker platform to monitor their performance. Don’t forget to include “more natural” language in your long-tails, either. Don’t be afraid to experiment. You’ll find opportunities to present themselves to enhance your relevance.

If your traffic remains the same, keep an eye on engagement. If the traffic and search queries are relatively static, that doesn’t mean your client’s site isn’t ranking for more relevant queries and driving high-quality traffic. Once you can pinpoint your trajectory, you can then optimize based on the simple rules I’ve outlined.

What’s the Bottom Line with BERT?

Well, that’s an excellent question. Any change, especially one as fundamental and far-reaching as BERT, will have major impacts in several areas for many years to come. The bottom line is that keywords no longer have to be as specific as they once were. The order, or lack of order, also now appears to be of less importance. This is a good thing for your business and us technophiles in the website content world — if you optimize for the update! Because these algorithms are designed to understand the language in a way that is, by its very nature, multicultural, then comprehensive SEO strategies will need to incorporate non-native speakers in their outreach. There’s a whole world out there of people speaking other languages in conjunction with English — and they are potential clients! 

While some so-called SEO experts are left twiddling their thumbs, this is actually a great time to take advantage of BERT’s emergence! True SEO pros stay on top of the latest trends and algorithm updates — and we understand that they can be put to immediate good use! There’s no time like the present to fine-tune your keyword research strategy for maximum performance in today’s BERT-centric online world.