Google BERT -
a natural evolution of Google’s algorithm

In late October 2019, Google announced an advancement to how they process search queries, utilizing a model called BERT (Bidirectional Encoder Representations from Transformers).
 
Google has been trialing and fine-tuning use of BERT since November 2018, however officially announced its full rollout in October 2019.
 
Whilst its initial launch was purely for English language queries, Google announced on December 9th that it would be rolling out to over 70 languages.

What is BERT?

In tech-speak, BERT is a ‘neural network-based technique for natural language processing (NLP) pre-training.
In simpler terms, it is a technique devised to help bots process language in a manner much more akin to how humans do. It is a method aimed at helping Google’s algorithm understand searches more naturally rather than a ‘bunch of words.

Google confirmed that this method will be applied to both search results and featured snippets.

Resolution's POV

The inclusion of BERT to how Google processes search queries can only be seen as a positive and a natural evolution of Google’s algorithm. For over a year we have focused our SEO recommendations on ensuring that content is worded naturally – partly due to the impact Voice search is having on how users search and partly due to the rise of featured snippets.
 
During one of his Webmaster Hangouts, Google’s John Mueller was asked what SEOs should be focusing on to align to the BERT algorithms.
In short, he explained that BERT is primarily focused around better understanding the questions or queries submitted in Google, and as SEOs we can exert very little influence on how users search.

So what can marketers do?

The examples that Google has predominantly given to show the potential change to search engine results are all focused on long-tail searches.

Long-tail searches – whilst low volume – tend to have high intent and are worded more conversationally.
As such, you should ensure that when creating content, you have conducted thorough long-tail keyword research and put your findings in to your content. Aim to write naturally and have your tone be conversational.

Other tips
 
If you are already tracking a list of long-tail keywords, be sure to keep a close eye on a monthly basis to your ranking performance – should you see noticeable drops it would suggest that the pages on your site relevant to these keywords need a content refresh.

Example: 

Google has provided some hypothetical examples as to how this update may evolve search results;
Bert-inner-image.png
 

For the BEFORE search result, Google’s algorithm has missed the importance in this query of ‘for someone’ and instead focused on what the algorithm has understood to be important. In the AFTER example it has now recognized the intent that ‘for someone’ adds to this search and provided a search result that correctly answers the intent.
 

Summary

Whilst our recommendation is to continue adhering to best-practice, it doesn’t downplay the significance of this update. It is considered the largest change to how Google processes searches since RankBrain was launched in 2015.
 
As BERT evolves (by its very nature it is always learning), we may start to see gradual changes in search results over time, but we do not expect to see any seismic differences in ranking volatility.
 
As with any significant Google update we will keep a close eye on organic performance and the studies that naturally fall out of these updates.
 
Our conclusion put simply, write content for users not search engines.


About the Author:

Gavin McColl is SEO Account Director at Resolution Australia - Melbourne. You can find Gavin on LinkedIn.

Call
Contact