In October 2019, Google announced one of the most significant advances in how it understood language since Hummingbird. The update was based on a natural language processing model called BERT, which stands for Bidirectional Encoder Representations from Transformers. The technical name is a mouthful, but the practical implications for anyone creating content for the web are straightforward and worth understanding.
What BERT actually does
Previous improvements to Google’s language understanding had made the algorithm better at grasping the general topic of a search query. BERT went further by helping Google understand the relationship between words within a sentence, and specifically how small words like “for,” “to,” “not,” and “without” change the meaning of a query significantly.
Take a search like “can you get medicine for someone at a pharmacy.” Before BERT, Google might have focused on the words “medicine” and “pharmacy” and returned results about pharmacies generally. BERT helped Google understand that the key part of that query was “for someone,” meaning the person searching wants to pick up a prescription on behalf of another person. That is a meaningfully different intent, and it deserved a meaningfully different result.
The bidirectional aspect of BERT is important here. Previous models read text in one direction, either left to right or right to left. BERT reads in both directions simultaneously, giving it a much richer understanding of how each word in a sentence relates to every other word around it.
What BERT meant for content creators
The implications for how businesses should approach their website content were significant. BERT rewarded content written in clear, natural language that genuinely addressed the topic at hand. It became better at understanding nuance, context, and the specific intent behind a search.
Conversely, content that had been constructed primarily around keyword patterns rather than genuine usefulness became less likely to perform well. If a page had been written to match a keyword rather than to answer a question, BERT was better equipped to identify that mismatch.
Google was clear that BERT was not something website owners could optimise for directly. The guidance was the same as it had always been: write for people, not for search engines. BERT simply made Google better at enforcing that principle.
The broader significance of BERT
BERT also represented a shift in the underlying technology driving search. The model had originally been developed and open-sourced by Google’s research team, and its application to search marked the beginning of a period in which advanced language models would become central to how search engines work.
For business owners, the significance is less about the technical details and more about the direction of travel. Google has continued to invest heavily in understanding language more accurately. Every step in that direction makes genuine, well-written content more valuable and keyword-stuffed, low-quality content less viable.
What this means for your business
If your website content is written clearly, answers real questions, and reflects how your customers actually talk about the things they need, BERT was a positive development for you. If your content was built around keyword formulas rather than genuine usefulness, it was a prompt to reconsider.
The practical takeaway is simple. Write as you would speak to a customer. Explain things clearly. Answer questions thoroughly. Use the language your audience uses. BERT made Google better at recognising when content does those things well, and that remains the standard worth aiming for.
