AI & GEO
BERT (Google Algorithm)
Last updated
What is BERT (Google Algorithm)?
BERT (Bidirectional Encoder Representations from Transformers) is a natural language processing model developed by Google and applied to Google Search in October 2019, initially for about one in ten US English queries and later expanded to many languages. It processes words in relation to all other words in a sentence simultaneously (bidirectionally) rather than sequentially, enabling much more nuanced understanding of query meaning, especially for conversational or complex queries.
BERT is particularly effective at understanding prepositions and context-dependent meaning: in one of Google's launch examples, a query about picking up medicine "for someone" at a pharmacy was finally understood as being about collecting a prescription on another person's behalf. Its introduction shifted SEO further away from keyword-matching tactics and toward content that genuinely and precisely answers nuanced user questions.
Why it matters for SEO
BERT is not a penalty. It made Google better at understanding what long, conversational queries actually ask, so pages that precisely answer that question gained ground on pages that merely contained the right keywords. Content built around genuine question-answering and contextual precision benefited most.
Example
Google's own launch example was the query "2019 brazil traveler to usa need a visa". Before BERT, results focused on US citizens travelling to Brazil; with BERT, Google understood that "to" meant a Brazilian travelling to the US and returned the relevant visa information. Small words that older systems ignored can change which page is the best match.
Free tools for BERT (Google Algorithm)
Go further with BERT (Google Algorithm)
Put the theory to work
LazySEO researches keywords, writes SEO articles and publishes them to your site on a schedule.