On 9 December 2019, an official announcement appeared on Google's Twitter account announcing the implementation of the search engine's then latest algorithm update, BERT. Do you know what it consisted of? What did it change? What impact did it have on positioning? How BERT was received by the world SEO? Are we in for another Google update? You will find the answers to all these nagging questions in our new article!
BERT algorithm - definition
Before talking about the BERT update itself, it is also important to introduce a concept without which it would not exist. We are talking about NLP, or Natural Language ProcessingBERT was developed as a technology to help the search engine to better understand natural language. BERT, on the other hand, was developed as an enabling technology for this process, and an extension of the acronym is: Bidirectional Encoder Representations From Transformers. "Bidirectional" - i.e. bidirectionality is what indicates BERT's revolutionary nature right from the start. The linear decoding of single words typed into the search engine, which has characterised the Google engine up to now, is abandoned. It replaces it with an understanding of each individual word (also conjunctions, conjunctions) and the meaning relationships between words. BERT is an artificial intelligence that has learned contextual interpretation and natural language processing.
This all sounds very mysterious, so it is worth pointing out what the purpose of the algorithm update is in the first place. Namely, thanks to it, Google is able to better understand the context of statements, its tone, and consequently delivers even more relevant search results to its users. Until the update, Google's robots also had trouble, for example, understanding pronouns, partials or phrases formed from words that form compounds with a completely different sense than that of each of these words individually. According to the developers, this is what is so innovative and groundbreaking about this algorithm - it not only interprets all words individually, but also having regard to their relationship with each other. It is through this that the search engine is supposed to read almost perfectly what the user wants when typing in a particular phrase in the bar.
BERT Google and SEO - what has changed?
At the end of October 2019, Google started implementing the new algorithm for English-only searches. As we mentioned, in December of the same year, its implementation was also announced for around 70 other languages, including for the Polish language. Many professionals then began to wonder frantically - how (and if at all) can one prepare for the change? However, most came to the conclusion that it was impossible to defend against BERT, but also not necessary. It is an algorithm that is not designed to penalise sites. It simply serves improving the level of understanding and interpretation of key words by the search engine and provide users with the most precise and tailored search results possible. Polish SEO specialists also knew more or less what to expect from the update. After the implementation of the BERT algorithm for English, Google reported that it had an impact on the 10% enquiries in the USA.
The only thing that could be done was to wait for the algorithm to be fully implemented and then, after the update, to analyse the phrases and see which ones we had seen declines on. The next step is to understand and adapt to the new mechanism and form of results. It is also worth noting that some SEOs were not at all afraid of BERT - quite the opposite! They pointed to the numerous benefits of its introduction, including the fact that the search engine would better read freely written texts without having to 'artificially' stuff keywords. The approach to research has also changed somewhat - key phrases have been chosen more frequently in different variations or with connectors.
The area most strongly "affected" by BERT is long tail phrases (long tail). They are more complex, precise and therefore also convert better than general phrases. Thanks to the new algorithm, the meaning of a long phrase typed into the search bar is more accurately understood. As a result, pages positioned on long tail valuable traffic has increased.
What about after BERT? Google SMITH?
Google is relentless in its efforts to improve search engine performance and precision. Although it is still less than two years since the implementation of BERT, another algorithm has already appeared on the horizon. SMITH. No official release date for the new algorithm has been announced, but Google has released an article scoring the benefits of the new update. Some speculate that Google has already implemented SMITH during the December 2020 update. It should be mentioned that it is not being developed as a replacement to completely replace BERT, but only as its extension, supplement.
But what is SMITH itself and what changes does it herald? The Siamese Multi-depth Transformer-based Hierarchical (as this is an extension of the acronym), as opposed to BERT, is supposed to read the meaning of a phrase not just based on the words around it, but on the entire block of text! And this means quite a change in SEO activities, especially in SEO copywriting and content creation. This is because Google will be analysing not only sentences with key phrases, but looks at the content from a broader perspective. The source will be verified 'from cover to cover' and in this way Google will match the best results for users.
Consequences of Google BERT - executive summary
Algorithms that came before BERT -. Penguin and Panda - strongly shook the entire SEO industry. Therefore, when specialists heard that the new update would affect as many as 10% search results, the atmosphere in the environment became tense. However, the reality turned out to be much softer than predictions. BERT was implemented and the perceived changes and drops (if any) were not catastrophic. Why? Because what BERT has changed is not the robots' evaluation of the site, but the assessment of the question search engine user.
English 
Polish 
German