BERT

BERT is a language model and the search algorithm update Google rolled out in October 2019. The name stands for Bidirectional Encoder Representations from Transformers, and it describes the mechanic: the model reads a sentence in both directions at once, so small words like "to", "from" and "without" keep the meaning they carry for a human reader.

Before BERT, Google treated a query more like a bag of separate words. In a search such as "can you travel to the US without a visa", the word "without" is the entire point, and that class of function word was what BERT handled properly for the first time. The practical effect on SEO was the end of chasing exact word combinations. Pages that answer the need behind a query started beating pages that merely repeated the right terms.

BERT is also the point where the technology behind today's AI answers entered search. It was an early transformer model, the same architecture that later produced the generative models behind AI Overviews and ChatGPT, so the shift from matching words to interpreting language runs in a straight line from 2019 to now. Understanding-based ranking is the foundation of semantic SEO, and the guide to AI search picks up what it means for visibility on the newer answer surfaces.