Google has announced a massive algorithm overhaul that impacts 10% of search queries and will likely create a seismic shift in how executives and SEOs approach SEO to drive more qualified traffic to their websites.
Pandu Nayak, Vice President of Search at Google, announced that Google is starting to use an artificial intelligence system called BERT (Bidirectional Encoder Representations from Transformers) to better understand and supply relevant results for ambiguous, complex, and confusing queries.
But what exactly does this look like in practice? Nayak gave the example of the query “do estheticians stand a lot at work.” Pre-BERT, Google would have pulled out individual words from the query, like “stand,” and matched it to a page that included the word “stand.” But this doesn’t always produce the most relevant results. With BERT, Google now looks at the word “stand” in context, realizing that it refers to the physical demands of the job.
Photo Credit: Google
How does BERT do it? According to Jeff Dean, Google’s Senior Vice President of AI, Google’s engineers trained BERT by feeding it paragraphs where 10-15% of the words had been randomly deleted and making it guess what was missing. Now that BERT is a component of the search algorithm, it will have plenty of opportunities to practice, as billions of Google searches are performed every day — about 15% of which its AI has never seen.
BERT means better, more relevant results for searchers — even for their most complex queries. Google hopes this helps searchers “let go of some of [their] keyword-ese and search in a way that feels natural.”
BERT shouldn’t come as a surprise
From the very beginning, Google has always made it clear that they do not want keywords influencing how people create content for the web, and has been working on maturing its algorithm to stop rewarding content creators who use that strategy.
- In 2009, Google confirmed it didn’t use the “keywords” meta tag as a ranking factor, with some speculating it could even be a negative factor.
- In 2011, Google started moving over to secure search, meaning SEOs started losing their ability to see what keywords brought someone to their site and started seeing “(not provided)” instead.
- In 2012, Google announced the Knowledge Graph (where the phrase “things, not strings” originated)
- In 2013, Google announced the launch of Hummingbird, an overhaul of its algorithm that allowed them to better understand the meaning behind the words, rather than just matching the words in the query to words in documents.
- In 2015, Google announced RankBrain, a new machine learning component to their search algorithm that improved Google’s ability to serve relevant results to confusing or never-before-seen queries.
BERT is yet another addition in a long line of improvements Google has been making to reward relevant answers in favor of keyword-matched answers.
What does BERT mean for SEOs?
For years, Google search results have programmed us to search in a way that we think will get us the answers we’re looking for. To date, that hasn’t matched the way we’d actually ask the question if we were to ask a human. This produced lots of queries that sounded more like “running shoes mens” than more specific, conversational queries like “best running shoes for men with flat feet.”
Now that Google, via BERT, can provide relevant answers no matter how conversational our queries are, SEOs will likely see an even greater uptick in the diversity and length of queries. What does this mean for keyword research and tracking?
SEOs have historically relied on painstaking research to select keywords to track, based on a number of variables — keyword search volume, strategic products or topics, query competitiveness, and more. While this strategy of keyword research and tracking seems alluring on the surface, in reality, its returns often don’t justify the time and effort we put into it.
It’s time for SEOs to embrace what every other marketing channel has: the importance of user behavior and intent as it relates to real searcher queries.
With the advent of BERT, optimizing for and tracking specific keywords makes less sense than ever. But SEOs have been addicted to keyword research and tracking for years — how can we break that addiction?
Using Google Search Console to track searcher behavior
As luck would have it, Google has provided a solution — Google Search Console (GSC) Performance reports.
As Google has lessened the importance of individual keywords in favor of context and intent, and as searcher queries have become more nuanced and conversational as a byproduct, Google has been quietly building its own keyword research and tracking tool (that just so happens to also be an alternative to scraping, reducing the cost of serving those pesky SERP scrapers).
GSC’s Performance report is user behavior on tap, with real impression and click counts for every query and searcher information like device and location.
The best part?
The raw data is free, and if you integrate the GSC API, you’re not limited to the first 1,000 rows of your keyword data like you are in the native app.
Botify customers are enterprise companies with complex websites comprised of billions of web pages, meaning it’s not uncommon for our customers to get impressions and clicks from more than 1MM unique queries in a given year. That’s why we’ve integrated the GSC API into our platform to create Botify Keywords.
How Botify Keywords prepares you for the age of BERT
In an age where BERT can help searchers get relevant answers to even the most complex queries, SEOs need a solution that helps them understand how real searchers are looking for and finding their sites — selecting keywords to track first and then hoping those are exactly the keywords your audience uses is backwards and likely can’t keep up with the pace and scale of how your real audience is searching.
Botify Keywords, powered by the GSC API, helps you understand how people are looking for and finding your site, holistically. This helps not only SEO practitioners, but also executives, understand their audience’s behavior in a way that helps them make informed, high-impact decisions about their site (without worrying about having to pay per keyword).
With GSC data in Botify Keywords, you’re getting real keyword insights — keywords your audience is actually searching (not just keywords you picked to track), with granularity like location and device, and the ability to view each keyword in context (not only ranking URL, impressions, and clicks, but alongside your other SEO data like title, internal PageRank, and much more).
You can read more about our approach to enterprise SEO keyword tracking, or about specific use cases for your keyword data (like seeing trending keywords) in our article Content That Ranks for Real Searcher Queries.
BERT changes the way SEOs need to be thinking about their audience and SEO strategy. Tap into user intent by navigating real search queries, create content that’s relevant to that audience, and you’ll get real results!