Google BERT Update and What To Do With It

Google BERT Update and What To Do With It

Despite many industry specialists noticing a mild (yet still unnatural) SERPs movements for nearly two weeks, no one saw this change coming. Less than a month after the pre-announced core update, the Google BERT Update was made public on October 25th, putting to rest many marketers concerns over fluctuations in SEO ranking. 

Whilst many argue that the changes won’t greatly affect SEO, we have a different interpretation.

So, what is BERT, then?

According to Google, the BERT Update specialises in “understanding searches better than ever before”. After finding that 15% of daily inquiries had never been looked up before, this put pressure on Google engineers to adjust search results to better represent searched terms. The issue was not that the questions were difficult to answer – the challenge lies within them being awkwardly phrased and therefore difficult for Google to understand and respond accordingly to. 

Whilst Google opts to have human-like conversations with users, users struggle shaking off the former conviction that the search engine only understands “keyword-use” phrases. Despite the Panda and Penguin versions dating back a few years (marking the end of the keyword stuffing era), people aren’t naturally warming to searching on Google in a more conversational tone. Think of it this way – if we attempted to talk to one another the way we speak to Google, we would have serious problems communicating. 

The core of the BERT update is having Google put words into context better. For example, searching a phrase with a verb ‘stand’ could result in getting back results including ‘stand-off’, ‘stand-alone’, ’stand-by’ and such. Essentially, BERT will see Google focus on enhancing its own functionality. Therefore, no one will get actually penalised despite changes regarding SERPs positioning being implemented. 

Okay, but what does it do and how exactly will it affect searches?

To clear something up, it is worth saying that BERT didn’t come into existence overnight. It is the first NLP (Natural Language Processing) system with no assistance, and therefore it took part in a long training period. In its essence, BERT crawls through page after page of text, to read our language and understand the grouping of information that comes with natural human speech. The key training ground of BERT was Wikipedia and sites similar to it. An interesting fact worth noting is that Wikipedia isn’t optimised for search – while there are titles, descriptions and keywords, none of the content is built to meet SEO requirements. 

Essentially, before the Google BERT update, the search engine would read content left to right as a human would, aiming to understand the context of a sentence. However, while humans have no issue understanding when the word is a verb or not, it is a struggle for Google. Our previous example with the word ‘stand’ illustrates this the best. Although it is a verb, Google would group it with nouns such as ‘stand-off’ and ‘stand-alone’ simply because it, unlike us, couldn’t tell from the context whether it was a verb or a noun.

With BERT, Google is now able to read right to left, doubling its understanding of the context. There are actually multiple tasks it performs once a query is made. Here are the three main ones: 

  1. Removing certain phrases from the search inquiry, essentially ignoring them. Theoretically, Google checks all the words and when they are used together to understand the content better. What it’s trying to do is match sentences on web pages that might be the best result. What is the best result, though? That’s task 2.
  2. Google BERT incorporates Next Sentence Prediction (NSP) to get a better understanding of what you are talking about. When it finds the match of your inquiry, it will also analyse the next sentence for contextual placement.
  3. Google can now see that removing certain words from the query can get more results. It can also use the second sentence to deliver much better results. Therefore the third task is delivering the final findings to the user.

See a visual example here

Whilst the update is predicted to affect 10% of all queries, it will mostly focus on long-tail queries. Therefore it isn’t likely to greatly affect many websites’ SEO. 

Essentially, BERT is all about understanding user intent better.

Does that mean there’s no need to optimise for BERT?

Google themselves said there’s nothing to optimise for with BERT other than opting to write content for users. The majority of the SEO community took that in with pleasure, throwing arms in the air and claiming if Google officially said there was nothing for them to do, this is how it is. We beg to differ.

Google’s BERT update will have an impact on content marketing. And while stating nothing needs to be done differently, Google has also emphasised that content will be more important than ever. To keep it nice and short, your SEO optimisation efforts should all be put towards on-site SEO and content, focusing on natural language and long-tail keywords. 

The following are the steps we advise taking:

  • Perform a Google SERP analysis;
  • Create relevant & optimised content;
  • Discover new keywords and opportunities.

Doesn’t sound like anything new, does it?

Think back to the findings we have listed before. Especially the part about BERT paying extra focus to the second sentence, and how context matters even more than the keyword.

While the urge to focus on long-tail keywords had been haunting marketers for several years, the Google BERT update is here to tell us this is mandatory now. In other words, follow the usual steps but shift the focus from best-ranking keywords to long-tail ones.

Final word

Google BERT may not seem like a big deal, especially for the SEO crowd. However, this is the first step and a sign of the direction Google will be going towards. Beginning to rethink your on-site SEO strategy will keep you ahead of the game when the major updates kick in.