What was “The BERT Update” and how did It impact organic SEO?
The BERT update is a significant change to Google’s search algorithm that was rolled out in October 2019.
It was designed to better understand the natural language searches (NLP) used by users, and as a result, provide more relevant and useful results.
In order to rank well under the new algorithm, it’s important to focus on creating content marketing that is white hat, follows EAT, very helpful, well written and informative, rather than keyword-stuffed and sales-oriented.
So far, the feedback on the BERT update has been really positive, with many users finding that their search results are now much more relevant and useful.
This is good news for businesses that are invested in creating high-quality content marketing, as it means that their efforts are more likely to be rewarded with higher rankings and increased organic traffic.
If you’re looking to optimise your company website for the BERT update, then here are a few things to keep in mind:
Focus on creating useful, well written and informative content marketing
Make sure your content marketing is written by an expert, yet is also easy to understand
Write naturally
Structure your content in an easily digestible way
BERT + Machine learning
SEO (Search Engine Optimisation) is the process of improving the visibility and ranking of a website or web page in search engine results pages (SERPs). SEO can be done through optimising the website content marketing, onsite seo, as well as developing off-site SEO tactics like link building and social media engagement.
How did Google build BERT??
Google’s BERT model is one of the most exciting recent advances in natural language processing (NLP).
BERT is a ” bidirectional encoder representations from transformers” – in other words, it’s a neural network that understands language by looking at text in both directions. This makes it very effective at tasks like question answering and sentiment analysis.
So how did Google build this amazing machine learning model? Let’s take a look!
They took a large dataset of English sentences (around 3 billion words) and used a technique called Masked Language Modelling to train the model. This means that they randomly masked out certain words in each sentence, and then asked the model to predict what those words were. The model was not only able to learn the general structure of language, but also to develop a deep understanding of the meaning of each word.