It’s been described as Google’s most important algorithm change in years – but what does the BERT update really mean? Ionie Ince, Head of SEO at We Influence, investigates.
Google’s Pandu Nayak (Fellow and VP of Search) announced the rollout of a huge change to Google Search starting in the US on the 25th October 2019 and, naturally, marketers went nuts.
So, what fresh horror awaits search this Halloween season? Well, it’s called BERT and it’s more of a treat than a trick.
What is BERT?
Simply put, BERT (Bidirectional Encoder Representations from Transformers) is an extremely advanced piece of technology that allows anyone to train their own Q&A system.
BERT has actually been knocking around since 2018 and was the result of Google pushing their natural language processing (NLP) capabilities but it’s only now that it’s being rolled out to search.
NLP is all about training machines to understand language the way that humans do. Whereas humans can pick up on colloquialisms (regional sayings, slang, etc.) and context, machines can’t. Or at least they couldn’t. So, in order to make machines better at understanding humans, they need to be trained, which is where BERT steps in.
What do you train machines with? Well, in BERT’s case, Wikipedia.
However, the special thing about BERT is the ‘bidirectional’ bit. Essentially, what this means is that BERT can better understand the context of all words within a query and the way in which they relate to each other. The results? A much more relevant answer to your question.
How does BERT affect SEO?
For SEOs, search is about ensuring that a client’s site ranks for the right keywords, meaning it will be seen by users who conduct searches using those keywords. Simple stuff.
BERT makes it even more likely that your site will be seen by the right people because it understands their search query better than before. This is a win-win as users find hyper-relevant information and sites get hyper-relevant traffic which is more likely to convert.
The effect of BERT being rolled out is going to be most keenly felt for sites that deal with longtail, conversational queries. Like news sites, for instance. BERT has also been effectively rolled out to all languages that currently support rich snippets (the highlighted results that appear in position zero). Didn’t notice? Good, that means that you’re correctly optimising for user intent (the purpose of their search).
If you’ve been involved in SEO for a while then you may remember Google’s Hummingbird (August 2013) and RankBrain (October 2015) announcements. Both of these seem like natural predecessors to BERT as they both altered the way in which SEOs optimise for longtail, conversational search.
As a reaction to these updates, SEOs began to create content which targeted query-based keywords (how, what, why, etc.) which caught users at different stages of the marketing funnel. Subsequently, the importance that was placed on this content and it’s quality seemed to feed into Google’s Medic update (August 2018). Now, with BERT, we’re seeing a refinement of search results based on improved understanding within Google’s core algorithm.
How can I optimise for BERT?
You can’t. BERT isn’t here to tank your site’s rankings; it’s not punishing you. Instead, BERT is here to help the right users find your site and if you’re committed to creating useful, high-quality content you should be in a strong position.