Skip to main content
News & Perspectives

Google BERT: What this SEO algorithm update means for the future of search

On October 25, 2019, Google announced a large algorithm update widely known as BERT. This update affects “one in ten search queries” and Google is calling it “one of the biggest leaps forward in the history of Search.” There is no need to panic or overhaul your search strategy, and you likely won’t see any major swings in traffic or rankings, but you should definitely take notice. This algorithm update means big improvements in the search landscape.

What is the Google BERT update?

Formally, BERT is “a neural network-based technique for natural language processing (NLP) pre-training called Bidirectional Encoder Representations from Transformers,” but let’s break it down.

BERT is a project that began a year ago in late October 2018, that focuses on processing keywords all together in a cohesive strand rather than one-by-one. The result is a search engine results page (SERP) that better matches the intent of the user’s query. For instance, in a search like “brownie recipe no nuts,” Google previously might have seen “brownie” and “nuts” as key terms and provided recipes for brownies with nuts. After BERT, however, Google better understands the “no” modifier and provides more relevant results. Below are some examples of results you might have found for the search before and after BERT.

Before

Search result for brownie recipe found before BERT update

After


Search result for brownie recipe found after BERT update

In the Before example, “no” is included in the search, but Google presents it in the wrong context, as “no-fail” and “no nuts” have vastly different meanings. In the After example, “without nuts” appears bolded, matching the intent of the search query “no nuts.”

Google has shared some additional examples of search results before and after the BERT update in their release statement.

What does BERT affect?

The current release of the BERT SEO update will affect organic results of 10% of English queries in the United States. Google will eventually roll BERT out to include more languages and locations. Google is already utilizing a BERT model to improve featured snippet results in two dozen countries where this SERP feature is available.

What does this update mean for the future of search?

Google is constantly evolving to provide the best experience for people, whether on the web or through spoken language. We’ve seen Google make some impressive strides with natural language processing in the past couple of years. According to our own research on voice search, Google Home is the best voice assistant in terms of user comprehension and results. In addition, Google absolutely wowed everyone with the May 2018 release of Google Duplex, the A.I. assistant that makes calls on a user’s behalf. During the demo, Google played a real recording of the A.I. assistant making a haircut appointment and the audience marveled at its ability to engage in a fluid conversation with a human with perfect comprehension and appropriate responses. The BERT model will only help to further improve the connection between humans and technology. 

Better translations online and in-person

Google experienced a breakthrough with BERT when they began researching Transformer and how well it is suited for language understanding and translations. Instead of translating each word on its own, the Transformer takes every word in the sentence into account, and decodes the intended meaning, generating each word as it relates to the context. With the release of BERT, we should expect to see an improvement in Google powered translations, including Chrome-translated webpages. This helps break language barriers across the world, and brings previously inaccessible web content to a larger audience.

Google has been working on expanding its language capabilities on devices since real-time translation built in to the Google Pixel Buds in November 2017 and continued progress with its release of Google Assistant’s multilingual capabilities in August 2018. Most recently, Google released Interpreter Mode for the Google assistant in February 2019, which expands upon the real-time translation capabilities, supporting 29 languages across multiple devices, but the translation service still has flaws. Utilizing the BERT model, Google should be able to vastly improve these real-time translations and help make cross-language communication a little bit easier.

More improvements in voice search

Voice search has the potential to disrupt the way we search, but it has yet to take off in a large way. The current use of voice assistants is widespread but simple, with more people utilizing voice agents to complete simple tasks such as playing music, setting alarms, and getting directions rather than searching the web. In our voice search survey, we found that some of the barriers to taking the next step with voice search capabilities include lack of assistant comprehension, poor search results in response to user queries. With improvements to natural language processing, we should expect to see a greater level of use of voice search assistants and more advancements in SEO for voice.

How do I optimize for BERT?

At this time, you don’t need to change anything about your current organic optimization efforts specifically for the BERT update. Google’s top goal is always to provide the most relevant results for each query, and as long as you continue to create content and web experiences with a user-first mentality, you’re already optimizing for BERT. Looking to the future, however, you may want to optimize your site for voice search, make sure your language tags are correctly implemented, and pay closer attention to the long-tail queries driving traffic to your site. We’ve reached the next level in search technology, and artificial intelligence is going to be a huge part of it.

— 

Contact us to learn more about the human-centered future of voice search strategy.