Breaking Down Google’s New BERT Update

Posted on December 20th, 2019 by nms

In an effort to refine the user experience and to improve search results, Google continually updates its search engine. Updates and revisions to the core algorithm occur on a regular basis. Google has even incorporated artificial intelligence and machine learning technologies into its search product, providing users with relevant and insightful results for nearly every search query. A recent addition to the Google search landscape is the BERT update. In this guide, we’ll explore BERT, how it works, and how it can influence insurance agency SEO (search engine optimization). 

What is BERT?

BERT is the acronym for “Bidirectional Encoder Representations from Transformers”. In simple terms, it is a machine-learning technology that gives Google’s search engine the ability to process natural language queries. BERT uses all the words in a given search sentence; an example is “affordable homeowners insurance agencies near me.” The technology looks at the context of each word and its relation to the words before or after it. Unlike previous natural language processing (NLP) technologies, BERT can understand word context in both directions, hence the “bidirectional” portion of its name. 

BERT was trained on the text of Wikipedia and used an unsupervised neural network to gain a better understanding of complex sentences, including the context of how each word in a sentence was related. 

Why is BERT Important?

Just as the Google search user experience has evolved over time, so to have user habits. At the beginning of online search, users would input a single search term or a simple string of words. Today, especially with the adoption of voice-based search tools on mobile devices, users are more likely to speak or type in full sentences. These are sometimes referred to as “long-tail search queries.” Prior to the adoption of NLP technologies, Google’s search algorithm had a difficult time processing sentences or long phrases, often misinterpreting user queries and delivering less-than-relevant results. 

BERT ultimately improves the search experience by attempting to understand what the user is after. By evaluating all the words in a given search query and how they relate to one another, search results are more accurate, more relevant, and more informative. 

Limitations of BERT

Google itself said that BERT is not yet perfect, and may not always fully understand the context of a given search query. According to MOZ, a leading online search information resource, BERT has a particularly difficult time in processing what is known as negation diagnostics, or what things are not. In other words, BERT may not accurately predict the answer when posed with a search query like “professional liability insurance is not…” 

There are two other primary limitations of the BERT technology. First, it is currently available only for English-language search queries. Second, it will only impact about one out of every 10 searches on the platform, influencing results that rank for those search queries. For users in other languages, Google expects to roll out BERT functionality in the coming months; it is currently training (remember, it is a machine-learning technology and is always learning from its experiences) on English-language searches and will use that training to better understand queries in other world languages

How Can Web Content be Optimized for BERT?

When Google first announced the rollout of BERT and how the technology worked, the first thought on many SEO professionals’ minds was, “how can we optimize web content to take advantage of this new algorithm?”

The answer is simple: one can’t. Because the algorithm looks at word relations and context in sentences, the only way to truly prepare web content for indexing under this new algorithm is to make sure such content is well-written. For insurance agency SEO purposes, content developers need to gauge what potential site visitors are looking for in terms of information, and then create pages that help inform those visitors. Well-written content is:

  • Clear and concise
  • Filled with details
  • Free of grammatical and spelling errors
  • Arranged in a logical structure (introduction, body paragraphs, conclusion)

By applying these principles to all web content, insurance agency SEO professionals can help improve search visibility. With great content available on a given blog or website, Google’s search algorithm can better understand the context of content and can then deliver more accurate results to users. 

About Agency Tsunami

Agency Tsunami is the digital marketing division of Neilson Marketing Services, who has provided the insurance distribution sector with innovative marketing and communication solutions for over twenty-five years. With our Agency Tsunami program, we have implemented an integrated marketing strategy that includes optimized, professional websites supported by high-quality SEO and Social Media for hundreds of insurance retailers and wholesalers. Our goal is to help your insurance business perform and generate new opportunities. Give us a call at (800) 736-9741 to learn more about how we can do that for you.