In October, Google announced the biggest change to its search in five years: Bidirectional Encoder Representations from Transformers (BERT). BERT is designed to better understand nuances of words within search terms and uses some heavy-duty neuro-linguistic programming, as well as Google-improved cloud-based computing, to cope with the processing. As with any such change, rollout will be in stages, and the full impact will be felt in 2020.

By January 2020 BERT was rolled out across 70 languages and is already being used to better understand content on your site and will inform how Google chooses to surface that in their results.

A good example of how BERT works is the query “2019 brazil traveller to USA need a visa” This is about a Brazilian traveling to the U.S., and not the other way around. Previously, Google algorithms wouldn't understand the importance of this connection, and returned results about U.S. citizens traveling to Brazil. 

Due to its desire to improve search results, Google estimated that 10% of search results would be affected and most of this would be longer tail search terms. The type of content that affected may have been poorly written, using templates, badly translated or written purely for SEO purposes. If you did see a drop-in traffic the 'value' of that traffic was likely to be low and poorly converting.

As with many updates that Google makes to their algorithm there is a temptation to quickly try and find out how it works and what we can do to take advantage of the changes. But BERT is nuanced, ever learning and changing, focussing on understanding how sentences and phrases should be interpreted