What is Google's BERT?

The AI That Learns How People Speak

What is BERT? Everyone is talking about how BERT is changing the way we search, the way bloggers like myself write their blogs, and the future of AI logic. Today we are going to explain what BERT is and how it's a great change to how search engines work.

If  keep an eye on tech news, you might have heard of BERT. October 25th, 2019 Google made an epic announcement about their new AI-powered search algorithm. It might have you confused and wondering how this is going to affect you.

Not to worry, we break that down and help you to see how BERT can be a friendly update.

What is BERT?
Thanks Sesame Street for not only being a great teacher, but giving us nerd role models. We love you!

For the first time since 2016, Google has done a massive update that changes the way that they rank websites.

What Is BERT?

BERT ( (Bidirectional Encoder Representations from Transformers) is a learning algorithm that tries to help direct searching with everyday language. That means that this AI is learning how people text, talk and ask questions. BERT reads websites like a human would to provide a better connection between the search engine and the readers.

How many times have you wanted to look up a recipe for Butternut Squash Bisque by typing in “Butternut squash bis”  and you had a spelling challenge and forgotten how to spell bisque.  But you hope that Google will just know.  Then the search comes back with a bunch of queries that have nothing to do with making delicious soup!

BERT is here to help with that.

By adapting to recognize full sentences, BERT is basically learning how to read by looking at the words around the sentence to understand.

This blog uses affiliate links to keep the costs down and this blog free to use. If you make a purchase through those links, we will receive a commission at no cost to you.

What we know for sure:

  • BERT is still learning and adapting to how we humans write and how we search.
  • Even Google isn't sure how this will affect SEO
  • This will mainly hurt blogs with spammy content.

Let’s talk about what BERT is looking for.

One of the better uses for BERT is through the new wave of voice searches. Voice search is a hot topic and BERT is supposed to help with connecting our voice searching to better answers because it understands basic speech and language.

BERT uses natural language processing to process the information as well as looking at your request from both directions. 

I use my Echo Show in my kitchen every day and ask Alexa to look for recipes.  Or I’ll grab my phone and ask Siri to give me directions for getting to Target when I’m traveling and I’ve forgotten something. Both times I don’t put in keywords, I ask a basic question.

BERT loves Natural Language. It is trained to look for information from both the left and right for the context of the sentence to try and understand what you are asking.

How to use BERT

How can I work with BERT in my blog?

BERT wants you to create hyper-relevant content that is relevant to search intent.  This means it is looking for the “Why” behind your search.

This AI loves Longtail keywords, but when they are optimized to answer questions, give complete sentences, and aligned more with what people are actually searching for.

The Types Of Searches: Education, Navigation, Transaction, and Review

The primary reasons people search are to learn something. So your content needs to address what people are looking to learn and teach it to them!

People also use search as a navigational tool.  They are looking for something specific like a product or a tool.

Transaction searches are for people who already know what they want, they now want to know where to buy and how to get coupon codes or discounts.

And Reviews for products or services also are a type of search that customers are looking for to give them insight into their transactions, this is called a commercial investigation.

How can I get BERT’s attention?

BERT wants you to tell it which type of post you are writing.  

Informational posts need to address the basics – Who, What, When, Where, and Why.  

“What is the best way to talk to BERT?”

“What recipes should I use for my holiday party?”

If you are writing about a specific website or product, you want to use the brand name or website. This is called a Navigation search.

“NerdMama is a tech blog for nerds and non-technical people”

Your reviews fall under Commercial investigation, so make sure you tell BERT you are writing reviews for the brand product.

“NerdMama’s review of Amazon’s Echo Show 2nd Generation”

And lastly if you offer discount codes or special pricing you want to grab that transaction search by calling that out.

“Our Joy Organics 15% off discount on 100% THC-Free CBD products is Firefly ( That’s a real code!)

Here are the best places to talk to BERT: Let’s use the topic of writing a blog post.

Headlines – This is the best place to tell BERT what type of post you are writing and the readers will know if you have the best post for their needs.

“5 Easy Tips For Creating Better Blog Posts”

In the Headers-make sure you are answering the questions.

“How do I write a blog post that everyone wants to read?”

Don’t forget the meta descriptions!

This is the best place to tell the search engines the point of your post and the keywords you are looking to address.

I use voice search all the time in my day to day life.

Even though this update might seem scary at first, I love that this means that using regular speech or asking questions, will result in better answers.

What this means for bloggers, like myself, is that we can go back to writing for real people instead of writing for keywords. Tell your story, write like you have something to say. BERT wants to hear it!

Have you ever been punished by a Google Algorithm update?  What happened?

We want to hear your stories about what the updates have done for your blog and what did you do to change that?  Leave a comment below OR join us on Facebook to talk about it!

This site uses Akismet to reduce spam. Learn how your comment data is processed.