On October 25, Google’s official company blog read that an update based on neural network technologies and thus on artificial intelligence went online: BERT. It should affect around 10% of all English-language searches. Google calls it the biggest update in the last five years – and a big step forward in the history of Internet search engines.
Since this publication, website operators and marketing experts have been dealing with a topic around the world: Does BERT influence search engine optimization? We took a closer look at the topic for you.
What is the Google BERT update?
BERT stands for “Bidirectional Encoder Representations from Transformers”, a technology developed by Google in the field of deep learning. It was published as an open-source in the past years. Basically, Google has revised its algorithm so that the update tries to better understand human language with all its complexity – the processing of natural language (NLP = Natural Language Processing). It is designed to improve machine recognition of natural language. Words are now considered in connection with other words from the search query so that Google can better understand the context of the search query.
So far, Google has looked at each word individually, one after the other, now it is checked at which position the word is located and in what relationship the word is to other words, e.g. through prepositions. The search engine is to learn to recognize word meanings within a sentence – including all the nuances that arise from the context.
With machine learning to more suitable search results
Including the context improves the quality of search results in two ways: on the one hand, the queries are processed more precisely and on the other hand, websites can also be evaluated more comprehensively.
Users do not always put their search queries in the form of pure keywords. In most cases, we use more than a single search term and more and more often also “naturally” formulated questions – or at least the versions of them reduced to the most important keywords. Words such as “for” or “without” may appear, which have so far had little influence on the search results. However, if they are used to build a relationship between the keywords used, many search results previously displayed as possible can be excluded.
In addition, context processing makes it easier to find keywords that have different meanings in different contexts. If the search engine learns to recognize these ambiguities and to infer the desired interpretation from the context, it can process search queries with a high degree of complexity all the more reliably.
BERT Update and SEO
Voice Search, in particular, will benefit from a refinement of language understanding by search engines. Search queries that we do not write, but speak, are much more likely to have naturally formulated questions and so-called long-tail keywords than is the case with “classic” typed search queries. And these long-tail keywords are characterized by the fact that they do not consist of a single word, but of a certain combination of terms. The relation of the individual keywords to each other determines which search results match the query – and thanks to BERT, the search engine learns to determine these relations as precisely as possible.
How can websites be optimized for BERT?
From an SEO perspective, there are no concrete conclusions that could be drawn for your own website. Ultimately, the update aims to understand more complex search queries and to provide informative search results. From this, it can be concluded that SEOs must continue to focus on the fact that the content is informative and offers added value for the searcher. If you can put yourself in your readers’ shoes and find out which questions are likely to lead to the search for the content of a website, you can both take up these questions in the text and give suitable answers.
The update is a further step towards creating the content for the user and not for the algorithm
Since BERT is able to process natural language, web texts do not have to be particularly structured, content-wise or stylistically revised in order to be able to be evaluated. A text that is comprehensible to readers is all that BERT needs for its ranking. The only factor that website operators should focus on in order to accommodate BERT a little is the intention of search engine users. However, this approach is neither new nor specifically tailored to BERT: it is basically the same strategy that is used to optimize Voice Search. In addition, the maintenance of a keyword density matched to the previous SEO conditions should not (yet) be neglected. BERT is currently not a replacement for the previous mechanisms, but an addition. The keyword frequency, therefore, continues to have a major impact on the ranking of a website.
If you want to keep your website up to date, you don’t have to do any special BERT SEO. Adequate keyword density and the use of long-tail keywords suitable for voice search will also be the focus in the foreseeable future in addition to further SEO measures.
How does the update affect rankings & which search queries does it affect?
The update affects both the organic search results and the featured snippets, but only to a limited extent. According to Google, BERT would be used in every tenth search query from Google.com. The rollout for other countries will gradually take place, still undefined. However, the featured snippets are already based on the BERT update in all 25 languages. The rankings will change in the sense of the BERT update, especially for search queries where the context has so far been left out or misinterpreted.
This primarily affects long-tail search queries and less those search queries that only consist of a small phrase. Google’s example of this is, for example, the search query “2019 brazil traveler to the USA needs a visa”. So far, Google has also shown search results that explain how US citizens can travel to Brazil, but the context of the search query means exactly the opposite. This is exactly what BERT can now interpret and play out other results.
How important do you think the update is?
The update is the biggest algorithm change in recent years. Since Google strives to make the search results as informative and tailored to the search query, it is essential that Google learns a contextual understanding. Search requests are getting longer, there are already around ? in the long-tail area. This means that search queries are increasingly adapting to the form of linguistic questions and no longer consist of just a few keywords. Users want to get an even more precise answer with specific search queries.
An evaluation of the search query without understanding the context would only result in unsatisfactory SERPs. This is why this update is important for Google as well as for users since they can expect even more coherent search results.