As with any Google algorithm update, when BERT was released, the SEO community were awash with excitement. Online marketers are speculating as to how the update works and what SEO strategies can be used to optimise for BERT.
That's why we wrote this article. We want our clients and other Shopify merchants to understand what BERT is and if it actually affects their approach to SEO.
What is BERT?
BERT is an update Google made to the inner workings of their search system around October 25th 2019. Google describes BERT as a 'neural network-based technique for natural language processing'. Its actual name is 'Bidirectional Encoder Representations from Transformers', but everyone calls it BERT for short.
What does BERT do?
BERT is an advancement in Google's natural language processing capabilities. It's not like previous updates such as Penguin, Panda and others which were designed to reward and demote the rankings of certain sites.
BERT has similarities to Rankbrain in that the changes take place at Google's end. What BERT does is improve Google's ability to understand the true intent behind a search query. It's a massive machine learning model designed to help Google’s systems better understand English language.
The core feature in what makes BERT special is how it can understand the relationship between different words in a passage of text. Other language processing systems process words one-by-one in a single direction. BERT can understand the context of a word by looking at the words that appear before and after it.
As a very simplified example, consider these two statements:
*“The bow was tied”
“Shot from the bow”*
The common phrase in these statements is the word “bow”. The word ‘bow’ is a homonym, which means whilst it is spelt and pronounced the same, it has different meanings depending on the context in which it is used. In a passage of conversation, humans should understand the context of “bow’ in each of these statements, but it’s not so easy for a machine. BERT should be able to understand how the context of ‘bow’ is different in each of these statements by looking at the other words that appear near ‘bow’.
As conversational search grows, so does the way people search on Google. Voice searches tend to include pronouns such as 'it' and 'their'. Pronouns can be challenging for a search algorithm to process. Because of their ambiguous nature, it's not always obvious what entity the pronoun is referencing.
What BERT should do is return results more aligned with the needs of the searcher. Google will have a better understanding of the actual intent behind the search, and will then be able to retrieve more helpful results from the index.
Why BERT doesn't actually change SEO
When Google release algorithm updates there's normally an influx of guides claiming to know what the update involves and how you can optimise for it. But that's not the case with BERT. BERT shouldn't actually affect your approach to SEO.
We know from the search rater guidelines and previous updates how much value Google place in expertise, authority and trust. We believe BERT is a continuation of this philosophy.
Subject matter experts will reference semantically related words, phrases and idioms in a natural and authentic manner. They probably won't be thinking about keyword SEO focus. It's this sort of content that we believe BERT could reward. Not because it's better optimised for SEO, but because Google is better at recognising its context and quality.
If we were to recommend a strategy, it would be to pay closer attention to the quality of your written text. If you are over-optimising and trying to shoehorn extra keyword phrases into your content, it could have the opposite effect to what you're trying to achieve.
Rather than trying to make your page 'more' relevant, focus on making the purpose of your page more obvious to a search engine algorithm. Ambiguity is like an SEO weed. It confuses search engines and dilutes the SEO value of your pages.
Remember, BERT advances Google's ability to understand the true context of a word by looking at other words which appear in close proximity to it - both before and after. With this insight, you can see how redundant it is to focus on outdated metrics such as keyword density.
Structured data may play a role in BERT. Whilst structured data probably has no direct effect on rankings, it may help search engines understand the topical relevance of your pages. Structured data is another opportunity to remove ambiguity by defining exactly what your pages are.
Yes, you should still take influence from keyword research and yes, you should optimise the topical context of your content towards the theme of audience you want to attract. But this needs to be done in a natural and authentic manner.
Our Recommended Approach to SEO
Wrapping up, we’ll emphasise that there are no special BERT strategies you can take advantage of. Instead, we would focus on the recommendations below which we believe will bring about SEO success.
- Hire professional copywriters to create your written content
- Have your content reviewed by subject-matter experts to ensure authenticity and credibility
- Read your content out loud - does it sound natural, or does it read like it’s written for a search engine?
- Remove any ambiguity. Make it crystal clear to a search engine what the purpose and contextual relevance of your pages are
- Don't over-optimise for specific keywords and phrases
- Do make appropriate use of structured data
If you want to learn more about BERT, we’d recommend checking out Dawn Anderson’s fantastic summary of BERT and the foundations of natural language processing over on Search Engine Land.