October 28 2019

Introducing BERT: Google’s Latest Update

When thinking about how your search behaviors have changed over the years, you may notice that the number of words you use in each search has increased. Where you once typed “sheet metal” or “running shoes” into Google, you now look for “annealed 4130 steel sheeting” or “nike running shoes for high arches”.

Over the years, the trend of longer phrases and a more conversational style of search has become the norm. In fact, when reviewing a sample set of data from our clients’ analytics, we found that 84% of paid search queries included three or more words. The intent behind those searches is important too. Those longer tailed paid search queries produced 77% of the form submissions from our sample set. This is a great example of the search defining the user and understanding that the intent behind longer, more specific search queries is often greater than shorter, broader searches.

With its most recent algorithm update, also known as BERT, Google is attempting to make its results more relevant to a user’s intent by better answering their queries. BERT stands for Bidirectional Encoder Representations from Transformers, which is just a fancy way for Google to say that they are now focusing on how the words used in a search are connected to each other, and therefore attempting to create more relevant results.

BERT began being rolled out on October 25th and is expected to continue to impact search results over the next several weeks. Languages other than English will be rolled out later.

This update is not replacing RankBrain, the update from 2015 that was the start of Google’s exploration of how users use language to search. It is an extension of what RankBrain does, only now, instead of focusing on the main one or two words within a query, Google will look at how the words are connected to one another and what those relationships mean.

What Is Next

This update is expected to affect as much as 10% of all searches (most algorithm updates impact 1% or less), but it will be a subtle and refined change instead of an avalanche. Where once before, Google would have focused on just a couple of words in the search query, now they will look at the sequence and relationship of the query itself – particularly where relational words are used, such as “to” and “for.”

For example, take the search query “parking on a hill with no curb”. Before BERT, Google would have focused on “parking”, “hill” and “curb”, which would have yielded a more general result. Now, the word “no” would be recognized as important to the statement, and a more accurate answer would be given.

Image Credit: Google

One of the main impacts of this update will be on the featured snippets. As Google works to understand the intent of queries, featured snippets will appear with more relevant information. For those sites with detailed information and FAQs, this algorithm is likely to have some affect as well.

What to Do

There is nothing to do. Good, clear content was always important, and this update only reinforces that idea. Make sure your content is user-friendly, written clearly and answers common questions a user may have. By using TopSpot’s B-SMART Method®, you can ensure that you’re keeping the user in mind across all touchpoints within their purchasing cycle. This will help your rankings in Google, and also make your customers more likely to find you. As Google’s Danny Sullivan put it, “There’s nothing to optimize for with BERT, nor anything for anyone to be rethinking. The fundamentals of us seeking to reward great content remain unchanged.”

The release of BERT only underlines the point that Google is trying to make search results more relevant to the user by providing content from quality sources. The ability to identify intent and context from organic search queries is a huge step forward in that direction.

If you want to know how to see if your site is being affected by BERT, read through our post about how to tell when an algorithm affects your site.

Would you like to receive more information like this, sent directly to your inbox?

Sign up for our newsletter