What’s the Google BERT update? & Can You Optimize For It?

Updated January 10, 2023

The Newest Google Algorithm Update

Google has updated their search algorithm. Again. This time it’s called BERT. Ernie is nowhere to be seen. So what is Google BERT? What is natural language processing? And how does it affect your website’s visibility in organic search? These are all great questions, ones we’ll answer and provide context as to what’s on the SEO horizon.

What is the Google BERT update?

BERT (or Bidirectional Encoder Representations from Transformers if you prefer) is a natural language processing (NLP) technique created by Google to help it’s search engine better understand the content on a web page. NLP is used in a number of different tools already, like chatbots or Google translate. The BERT is the newest NLP application for search. 

How does the BERT machine learning model work?

The idea is that BERT will help Google determine when writing is naturally flowing; it looks for familiar patterns in text. The software then uses this information to help it understand the context of the words on a page.

For a simple example, let’s look at the structure of a basic sentence:

The cat sat on the mat.

BERT would help the search engine to understand that this sentence contains a subject (cat), verb (sat) and an object (mat). Once it has “learnt” this, it can then use the rules to understand other sentences, and their relationships to each other. So if BERT is then given another sentence:

The dog eats the fish.

It can work out that the subject of the sentence is the dog, the verb is eat, and the fish is an object by following the rules above. Then with this information it can look at the rest of the page (or website) and check if the content matches what it is expecting to read.

Bidirectionality 

As well as understanding single sentences it can also look at them in order. So for example, if we go back to the sentence above ‘The cat sat on the mat’. And then add another sentence:

The cat sat on the mat. The cat went to sleep.

Google can tell that these two sentences are talking about the same subject (cat), so it assumes that they are related. The example we’ve used here is very simple, but BERT can also understand more complicated ideas as is shown in this image from Google.

Here you can see that it knows that the words ‘milk’ and ‘store’ have an association. But ‘store’ and ‘penguins’ or ‘flightless’ do not. So how does it know this?

The algorithm has access to a corpus. You can think of a corpus as a kind of dictionary, or vocabulary, of words with their associations mapped out. So BERT will check for the word ‘store’, open that page in it’s corpus, and look at all of the other associated words. If there isn’t a match, then the two sentences are not related.

What are Google BERT algorithms?

When websites like ours are talking about Google BERT or Google BERT algorithms they are saying that Google has made a change to the ‘engine’ behind the search engine. And that Google is now using BERT as a way of understanding what is on your website. Changes like this happen semi-regularly and can create some upheaval in website rankings as Google crawls through sites and re assesses them.

Will BERT be the game changer in NLP?

It remains to be seen if this is a ‘great leap forward’ or just another small step on the road towards better AI understanding. So far the initial signs are promising. It seems that BERT has improved Google’s understanding of some complex searches. Over a third of Google searches are now four words or longer. As voice search grows in popularity, it seems logical that we will be searching for information online in more ‘natural’ ways than ever before. So evolutions like BERT will surely help keep searchers using Google happy.

Can You Optimize for BERT?

Yes and no. The idea behind BERT is that it recognizes naturally written, high-quality text. So the only way to optimize for BERT is by adding well-written content on your website. Outsourcing to a service that uses software to generate blog posts? We suggest you rethink that strategy.

One way to better optimize your content could be to use simpler or more direct sentences. This keeps your writing clear and easy to understand. This is not really an SEO optimization tip, it’s more of a copywriting tip. However, as search engines continue to evolve, we anticipate this type of copy structure to become the norm. The days of tricking Google by keyword stuffing are long gone. The future is now, and it’s all about quality writing.

So, If your site is already like this, then you don’t need to do anything. You may have even seen an uptick in your site’s SEO. Check the data in Search Console; you can see information about average position, click through rate, impressions up to 16 months. 

For BERT optimization, there’s no code to add to your site or new plugin to download. Just make sure you continue to craft and publish quality content that’s genuinely useful to your audience.

Bing BERT?

We’ve been talking about Google a lot in this article but they’re not the only search engine out there that’s using AI to improve the quality of their search results. Bing also has their own AI powered search tools, and their head of Search Frederic Dubut believes, as we’ve hinted at above, that you should be focused on creating content based around ‘intent’ rather than keywords. 

An Algorithm or Common Sense?

What he means by this is, your writing should be focused on matching a searcher’s intent. If they search for “New shoes”, don’t offer them socks, or even used shoes. Offer them new shoes! This has always been the bedrock of good SEO practice, but as with anything in life, people are always looking for shortcuts.

As AI technology improves hopefully this trend will continue, and SEO will be less about tricks and hacks, and more about good writing. Because the cleverer the robots get, the harder we’ll have to work to trick them.

Photo credit – Top: Maxpixel
Photo credit – Bottom: Pixabay