On 24th October, Google announced that it had added a new element to how it interprets searches, which would impact 10% of queries. Called BERT (Bidirectional Encoder Representations from Transformers), this new technology will help Google understand the natural way we speak and write, in order to better interpret search queries.
What does it do?
Simply, the introduction of BERT will enable Google to understand the context of search queries in order to deliver more relevant results for users.
Previously, Google has always had what SEO consultants refer to as "stop" words - that is, words that would help convey meaning to a human, but that search engine algorithms tend to ignore because they are so common that they slow down the delivery of results. These are common words like "the", "be" or "already" that are used so frequently across the web that it would be nearly impossible for Google to rank sites for them. However, understanding their use in a phrase can radically alter the meaning of the phrase.
If, say, you were targeting the keyword "red and black shirt", Google would likely see this as the same as targeting the keyword "red black shirt". The word "and" shows up so many times on so many pages across the internet, that Google knows that it's not the most important word in the phrase when trying to understand what words are relevant. Under BERT, rather than thinking that someone is searching for documents with the phrase "red and black shirt," or just "red black shirt" Google knows that you want to find an object that is called a shirt that has both the colours red and black.
As another example, if you were to ask a search engine "When can I legally drive backwards?", you would likely be asking under what circumstances you can operate a car in reverse. However, prior to incorporating BERT into its algorithm, Google might have ignored the word "backwards" and returned results telling you at what age you can legally drive, or what the driving test looks like.
Using BERT, Google would understand the intent of your search better and would return results explaining when it is and is not legal to drive in reverse.
Google gave the example of "2019 Brazil traveller to USA need a visa." Before BERT, Google would have ignored the word "to" in that search and returned results on .com for people travelling from the USA to Brazil. Now, Google will understand that the results need to focus on somebody travelling from Brazil to the United States. They provide this before and after image:
What does this mean for my SEO campaign?
BERT should mean that you get a higher quality of traffic to your website.
Google has explicitly said that there’s nothing to optimise for with BERT, which makes sense. If you are writing for your visitors, answering the questions they have, solving their pain points and generally explaining your products and services thoroughly, then BERT should help deliver visitors to the right pages on your website at the right stages in their buying journey.
BERT means that someone doing research is more likely to end up on a blog post or article, and somebody making a purchase is more likely to end up on a product page or service page.
In addition, the introduction of BERT will see improvements in the user experience of voice search. We already know that people search on voice in a different way to typing their queries into a search engine. Voice search users are more likely to employ more conversational phrases. With BERT, Google will now be better equipped to interpret voice search queries and so deliver better results.
While it is still a little early to say for definite, what you may find is that the number of visitors to your website drops, but that your bounce rate and conversion rate both improve as a result.
Is it perfect? No. Probably not yet. But Google has been working very hard to help searchers find what they need with the fewest clicks possible and this is a step towards that.
Get in touch
Need help with your SEO and keeping your website in line with Google best practice? Get in touch with our experts, call on 0161 402 3170 or email email@example.com.
Take a peek at our SEO case studies here.