On Oct. 25, 2019, Google announced what was – in their words – “the biggest leap forward in the past five years, and one of the biggest leaps forward in the history of Search.”
It’s name is BERT.
As apocalyptic as this update sounds, the change is a major advancement in Search capabilities and should help real people find real answers.
Here’s everything you need to know about BERT:
What is the BERT update?
BERT stands for Bidirectional Encoder Representations from Transformers, which basically just means Google can now understand how words in a sequence relate to one another rather than interpreting them one by one, in order.
Google can now understand how words in a sequence relate to one another rather than interpreting them one by one, in order.
BERT is predicted to affect 10% of Search queries, but mainly complex or conversational queries.
In other words, you may not notice a difference in your shorter, simpler keywords, since those are naturally already easy for Google to understand.
Why was BERT important?
When I Google something new, I don’t always know exactly how to phrase my query. So, I guess at what words I need for Google to understand my thought process.
Everyone does this.
“We might not know the right words to use, or how to spell something, because often times, we come to Search looking to learn – we don’t necessarily have the knowledge to begin with,” explained Pandu Nayak, Google Fellow and Vice President, Search, in the Google blog post that announced BERT.
Nayak calls this kind of guessing at what to search “keyword-ese” – kind of like this query: “Can you get medicine for someone pharmacy.”
I know what this person is asking. And the Google of a month ago almost gets it – but not quite. BERT helps Google understand how important the words “for someone” is to this query. Here’s the before and after BERT comparison from Google:
Here’s another good example from Google: “2019 brazil traveler to usa need a visa.” Before BERT, Google didn’t help this person much at all because it didn’t understand that they’re going from Brazil to the U.S.:
How do I monitor changes related to BERT?
Chances are, you won’t notice much of a change in your keyword rankings. BERT affects those longer, more conversational or complex queries, which:
- Probably aren’t the keywords you’re optimizing for.
- Are more difficult for keyword-tracking tools to monitor.
Many of the queries affected would probably be either full sentences, or long strings of keyword-ese, as Nayak described.
As a result, there really haven’t been major fluctuations in SERPs. Here’s the SEMrush Sensor, which monitors SERP volatility for the past 30 days:
SERP volatility has been normal or low.
For comparison, here’s what the SEMrush Sensor showed after the (unconfirmed) Maccabees update in late 2017:
Of course, I still wanted to see whether I could see any difference in keyword positions after the update, so I did a small experiment.
In SEMrush, I looked at all of Brafton’s organic search positions from Oct. 30 until Nov. 12 for keywords that consisted of more than four words (since shorter keywords probably wouldn’t be affected much, if at all).
Of the 2,031 keywords with more than four words with organic positions in that timeframe, 31 had fluctuated:
- 9 went up.
- 15 went down.
- 7 were new (nice!).
These were not high-performing keywords. Only two of them had positions on the first page, and only three drove any traffic.
But I don’t expect keywords like “scary numbers that actually work” or “obey your thirst sprite commercial” to bring a ton of traffic to Brafton.com anyway – plus it’s probably not our ideal audience, either. They both went up in SERP positioning, though.
How do I optimize for BERT?
This is the easy part. To optimize for BERT, optimize for real, live human beings.
The whole point of BERT is to understand how people actually speak. It’s natural language processing. Remember, this is the same technology that lets us speak to our phones, and lets our phones talk back.
So, if you’re already creating content that:
- Answers the questions your audience is asking …
- Uses the language your audience uses …
- Provides helpful, relevant information …
… then you’re doing the right thing. Just keep on creating valuable content for your audience.
More resources on BERT and Google’s update history:
- Content Marketing Weekly: Under the hood edition
- A complete (and actionable) Google update history timeline
- Understanding searches better than ever before by Pandu Nayak, Google Fellow and Vice President, Search
- A deep dive into BERT: How BERT launched a rocket into natural language understanding by Dawn Anderson at Search Engine Land