BERT: Google’s Latest Search Algorithm to Better Understand Natural Language
Google’s ability to deliver relevant searches has made another advancement with BERT, the latest machine-learning algorithm update to Google. Since RankBrain launched in 2015, Google has not made a language-based algorithm update until now. BERT officially launched in October 2019, but the algorithm has been in the works since at least 2018. David Clemen, VP of Digital Marketing at LSC Digital, shares insights into what BERT is, common misconceptions surrounding the update and best practices for keeping your SEO strong in relation to Google’s algorithm
Watch the full webinar below or download the presentation here:
What is BERT?
- BERT stands for Bidirectional, Encoder, Representations from Transformers.
- In short, BERT can help computers understand language more like humans.
Why is BERT a Big Deal?
- Unlike previous machine-learning programs like RankBrain, BERT can train language models based on a whole sentence instead of isolating specific keywords in the query.
- BERT focuses on how words interact in relation to one another. For example, smaller words in queries like “to” get overshadowed by more important keywords, but those smaller connective words indicate different intent to the searcher.
BERT in Use
- Pandu Nayak, Google Fellow and Vice President of Search provided examples of BERT in action on the official Google blog.
- In an example, Nayak used the query “Can you get medicine for someone pharmacy.” In a SERP before Google implemented BERT, the results show general information about filling a prescription. With BERT, Google better understands that “for someone” is a crucial part of the query and therefore provides results about friends and family members picking up prescriptions on behalf of someone else.
- Regarding featured snippets, BERT is acknowledging those same nuances to provide more accurate position zero results.
- For example, a search for “parking on a hill with no curb” previously ignored the importance of “no curb” and would provide results for simply parking on a curb. With BERT, that no longer happens.
What’s the Difference Between BERT and RankBrain?
- RankBrain adjusts results by looking at the current query in relation to past queries. RankBrain, as discussed in previous webinars, uses location data, behavioral data and content freshness to generate results based on in-the-moment keyword searches.
- BERT looks at the content before and after each word to understand the meaning and relevance to the query.
- BERT is, therefore, a continuation of RankBrain and not a replacement for it. This is a continuation of Google’s expansion into AI-related businesses, providing better results to more complex questions.
With the limited information circulating about BERT, there have been misconceptions about the significance of this change. Here are the facts:
- Thousands of algorithms layer together to create Google’s search algorithm. They all work together. The BERT algorithm is a language model.
- While BERT is a big deal, it’s not the biggest update of all time.
- No, you CANNOT optimize for BERT.
- BERT and EAT are separate, but they work together.
What is Google’s EAT Rating?
- EAT stands for Expertise, Authority and Trustworthiness.
- EAT is not an algorithm, but a guideline for Quality Raters to assign quality scores to search results. It’s used in honing the algorithm.
- EAT specifies high quality content and considers authorship, not just websites.
- Add author names and bios for all editorial content. Not only is this a great practice for writers, but according to EAT, a quality rating will follow the author’s name and not just the publishing site.
- Build up social media profiles and engage with influencers in your field to build your brand. This positions you as a leader in your industry, and helps interact with your audience.
- The technical stuff isn’t glamorous, but it’s important in telling search engines that your site is secure. SSL certificates will help establish your site as secure. Verified badges and logos will also signal a high quality rating.
BERT’s core functionality will not help or hinder your website’s SEO. The update is meant to provide searchers more accurate results. This sets the stage for better quality voice search results, and drives home a point from LSC Digital’s “Top SEO Recommendations” webinar: write how you speak, not for the search engine.
Interested in Google’s journey? Check out the webinar “What’s Up with Google” on LSC Digital’s YouTube channel.