Google BERT Update

Listen on SpotifyListen on iTunes

Previous Podcast

Last week, Google announced one of its largest updates to date: the Google BERT Update. No, this has nothing to do with Bert and Ernie from Sesame Street – BERT stands for Bidirectional Encoder Representations from Transformers. 

Sounds complicated, but essentially this update puts emphasis on search query context that will affect 10% of searches. This update is going to affect nearly everyone, so stick around as we explain what BERT means for the future of search.

Learn more as Jailyn and Morgan discuss the Google BERT Update in layman terms and examples. Tune in, or read the summarized transcript below.

Icebreaker

  1. How many queries does Google see on a daily basis?
    Answer: Google sees at least 3.5 billion (with a B!) search queries every day – that’s 40,000 search queries every second on average. Source: Internet Live Stats.
  2. What is machine learning?
    Answer: Machine learning is the scientific study of algorithms and statistical models that computer systems use to perform a specific task without using explicit instructions, relying on patterns and inference instead. Source: Wikipedia.
  3. What are the implications for voice search?
    Answer:  As more and more web traffic comes from voice search, which is expected to be 31.6% by 2020, Google and other search engines are racing to adapt the way search works to provide more relevant results, no matter how the query was made.

What Exactly is BERT?

As we mentioned earlier, BERT stands for “Bidirectional Encoder Representations from Transformers,” but we haven’t necessarily stated what it means or what it does. 

Essentially, it’s a mouthful of jargon for a new technique for natural language processing through a neural network, which takes parts of speech, entity tagging, and question-answering into consideration when deciphering and answering search queries.

This update will better help Google understand the context of queries to serve users efficiently with more relevant results. 

BERT In Action

Here are a few, good before and after examples we’ve seen floating around the internet

Example #1

This example shows the searcher types the query as “2019 brazil traveler to USA need a visa.” Now before the update, the results showed more information about U.S. citizen going to Brazil rather than what the searcher actually typed. With the Google BERT update in place, the new results show Tourism and Visitor information from the U.S. Embassy in Brazil – this is more of what the searcher wants to see. 

The update will also take into account prepositions, like “for” or “to.” These prepositions add a lot of context to searches.

Google BERT Update – Brazil Example

16% to 20% of queries that get asked every day have never been asked before. Source: Internet Live Stats

Example #2:

Before the update, Google match terms, therefore it would similar words have the same meaning, which we know they don’t. Search results didn’t take into the effect of homophones, which are two or more words that have the same pronunciation, but different meanings, origins, or spelling. 

In this example, the result before the update though that “do estheticians stand a lot at work” meant that they want information about estheticians alone, and it compared the information of two different estheticians. With the Google BERT update, we see that the results are more accurate with what the searcher wanted to know in the first place. 

Google BERT Update – Esthetician Example

What Should You Do Now?

Since Google makes thousands of changes to their algorithms every year, it’s impossible to specifically optimize your website to meet the needs and demands of each update. This update focuses on search queries, not results, so you couldn’t even prep your website if you tried.

One thing you can do for your website is to have great, relevant content written in a conversational and natural language. Content will be more important than ever in the battle for search relevance. Write your content in a way that is easy for humans to understand and learn from. Don’t be vague, but have a clear takeaway or CTA (call-to-action) in your content, and you’ll be in a good place.

Before you update your content, use Google Search Console or Ahref to check which search queries are being affected. Figure out if the same or different content or queries are ranking. From here, you can see how you need to update your website content and make adjustments. 

Machine learning is going to continue to become more and more complex and accurate in its results, therefore our world will be changing in big ways. BERT technology is a huge milestone for machine learning, so it will be interesting to see where things go from here.

Need SEO Assistance?

We love all things related to SEO, from creating titles and meta descriptions to actually making sure your website rankings and content are up to par. Let us know assistance your SEO needs today by conducting an audit on your website to find your strengths and weaknesses. 

Leave a Reply

Your email address will not be published. Required fields are marked *

Read Related Posts