Effects of the BERT update on content marketing

Staying up to date with trends in marketing is a priority for any marketer who wants to deliver real value and results. A new trend arouses a lot of curiosity. It also gives marketers time to take a pause and evaluate their approaches to marketing, and whether they are aligned with the new trend. 

Such is the curiosity that surrounded what Google called one of the major algorithm updates in the last 5 years, the BERT update. 

In this article, we explore this update – what it means, and its effects with regards to content marketing.

What is the BERT update

In order to understand this update, it is important to understand how Google interprets user queries.

A basic understanding of Search…

For a while now, Google’s focus has been striving to understand users’ search intent (why people search for something). There has been a little success with longer keyword phrases, which are commonly referred to as long-tail keywords. However, if a user keyed in a query in a language that is more ‘human and conversational’, the search engine would often show irrelevant results.

This is because, for a long time, Google has interpreted queries based on individual words and what comes before or after them. That is why all you needed to get your content to rank well on Google a few years ago was to use the correct keywords, in the title, headers, and at the beginning as well as the end of your content. 

However, with technological advancements in fields like Machine Learning, Natural Language Processing, and Artificial Intelligence, the search engine has been able to increasingly interpret search queries based on the semantic meaning between words. 

In comes BERT…

The BERT update is based on a technology that Google unveiled in 2018, which is rooted in training a system to process natural human language. 

BERT, Bidirectional Encoder Representations from Transformers in full, is meant to process queries within their context, as opposed to one word following the other. This way, users will be able to get even more relevant results from Search. 

This example from Google’s blog shows the difference in contextual meaning that has been brought about by BERT:

Source: https://www.blog.google/products/search/search-language-understanding-bert/

What you need to understand about BERT

Here are the main takeaways about this new update:

  1. It will apply to both search results and featured snippets.
  2. It is going to apply to a tenth of search queries (meaning that out of 10 queries, it will only be able to interpret one in context).
  3. It is first being rolled out in the US.
  4. It will affect queries in English
  5. Eventually, the update will be rolled out to other locations and languages across the world based on the ‘lessons’ learned from making improvements in English.
  6. In 12 countries, where featured snippets are available, BERT is being rolled out, and Google has reported significant improvements in languages like Hindi, Portuguese and Korean.
  7. BERT is definitely a preparation for voice search, which people are increasingly using. Voice search stats show that 30% of web browsers will not have a screen, and 50% of searches will be voice searches. 

How does BERT affect content marketing?

It goes without saying that the BERT update will have an effect on the content shown on search results and featured snippets. It will only be ‘friendly’ to relevant and specific content that solves a user’s problem in the shortest time possible as confirmed by this tweet by Danny Sullivan:

Let’s explore the ways in which content may be affected:

Thin content

Websites and webpages with thin content could see a drop in ranking. This is because Google will be looking to understand a user’s query in context and show them results that are relevant. It might be difficult to get context from a page that is image-heavy with extremely little text. 

Image alt tags

Closely related to thin content is images. Even though the human brain is highly visual, search engine algorithms still use text to understand what is relevant to the user’s search query. This could mean making image alt text more ‘human, and conversational’ in order to increase the chances of Google ‘understanding images’ as relevant to a user’s query. It does not mean that you, however, go on and spam your image alt tags with irrelevant long-tail keywords. 

More structured content

Since BERT will be affecting featured snippets too, structured content might see a rise in ranks. Google might be able to ‘understand’ structured content better than content that has no sense of structure. 


A user keys in a search query because they have something specific that they are looking for or need to achieve. One of the temptations with long-form content is to try and capture several intents in one. A user wants to get what they are looking for easily, and in the shortest time possible. 

Content marketers could borrow a leaf from Question and Answer sites like Reddit, Quora and Stack Overflow. These sites provide answers to specific questions. 

While it may be difficult to capture intent, marketers need to go out of their way to create content that answers specific questions, while taking context into consideration. 

An example might be a user who keys in ‘top universities in India’. Such a user might be a student who wants to join a university to undertake a specific degree. It might be more valuable to create content that talks about top universities based on degree categories. You may talk about top universities for someone looking to learn Computer Science or Business or even Mathematics. 

Preparation for voice

With voice search growing steadily, user search queries will definitely become more complex. While 15% of searches on Google have not been seen before, this number is highly likely to increase. Content marketing has to completely move to longer tail keywords, that are really close to human language. This is the only way Google is able to determine context. 

In conclusion

While BERT is going to see an increase in search results’ relevance, it is not yet 100% accurate. The best thing to do is to ensure that your content is relevant, specific and solves a user’s problem in the shortest time possible. BERT can only get better with time. Only quality content will withstand its test, and of course, the test of time.