Warning: this content is older than 365 days. It may be out of date and no longer relevant.

You Ask, I Answer: Testing for BERT?

Mitch asks, “Danny Sullivan says there’s nothing to optimize for BERT – ‘There’s nothing to optimize for with BERT, nor anything for anyone to be rethinking. The fundamentals of us seeking to reward great content remain unchanged.’ – is anyone actually testing for BERT?”

Danny is fundamentally 100% correct – if you’ve always been creating great, rich content that meets Google’s Expertise-Authority-Trust guidelines then literally nothing changes. The challenge for a lot of marketers is… they’re not. And a lot of the “scalability” tricks people use – like hiring the cheapest outsourced content farms or having the intern crank out swill, or copy-pasting lots of pieces from around the web – all those fall apart more. To put it another way, the great remains great, while the mediocre becomes bad in terms of search performance.

BERT also changes the game for a lot of low-performing marketers because it’s harder to game. The skills and tools needed to use it and work with it are a step higher than what the average SEO practitioner typically has available.

So, nothing changes for the best. Life gets worse for the rest.

We did a test yesterday on a key client page that had lost over 2,000 organic search visits year to date, and ran the process I outlined in the blog post. What we found was that there were a few key topics that weren’t in the page that were prominently in the ranking competitive pages, topics that made logical sense but were just omitted from the original piece (possibly for clarity). So, changes made – now we wait to see what happens.

Here’s a link to the Trust Insights blog post I wrote on BERT.

You Ask, I Answer: Testing for BERT?

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

In today’s episode, Mitch asks, Danny Sullivan over at Google says there’s nothing to optimize for BERT, nor for anyone, anything for anyone to be rethinking the fundamentals of seeking to reward great content remain unchanged.

And Mitchell’s question, Does anyone else actually testing for BERT? Danny is fundamentally correct.

If you’ve always been creating great content that is useful or helpful or entertaining, that people like that answers the searchers question, and that means Google’s EA t guidelines, expertise, authority and trust, then yeah, literally nothing changes.

You’ve been creating great content, you will continue to great create great content and Google will continue to funnel people to your website.

That doesn’t change.

The challenge for a lot of marketers is There are not, right.

A lot of marketing content isn’t very expertly written, isn’t high authority and isn’t trustworthy.

And so this, the algorithms and things that are are being used, will push people away from your content because it’s not as good as what else is out there.

That’s, that has been the case for every single one of Google’s algorithm changes over the years.

And the fact that you know, they’re constantly tuning in real time.

There’s just not a place in.

If you’re a marketer who wants to have high performing content and attract a lot of search traffic, there is no place for mediocre content.

And this certainly is no place for bad content.

When you read Google Search Quality Rating guidelines, one of the phrases that really sticks out and has stuck out for me, ever since I read it is there’s an entire category of stuff that’s not going to rank well, that says nothing wrong, but nothing special, right? Nothing wrong, but nothing special is a death sentence to your content period.

It doesn’t matter what era of Google algorithm we’re talking about.

It means your stuffs not great.

So the other aspect of this is that there have been a lot of marketers and a lot of search firms to that use.

They called scalability tricks, and I can scale to any industry.

And what they do fundamentally is just hire the cheapest outsourced content farms or guest authors or have the interns crank out swell or at worst copy paste paragraphs from you know, different sources around the web, to put together high volume content that’s low quality.

Anytime you’re dealing with a natural language processing algorithm, especially a deep neural network model, like BERT or XL net or GPT to order Elmo or glove, or whatever, the ones that can see beyond the center so word to phrase to sentence to the paragraph to document, right? I have absolutely no doubt in my mind that Google’s ability to use these algorithms extends beyond the sentence level to the document level.

And in fact, I believe if I’m not mistaken bills Lawsky over it, SEO by the sea has explored some of Google’s patents on this topic.

All those cheap tricks that people use just to crank out content.

That’s low quality break more, in when you start using more advanced neural and natural language processing algorithms, the more advanced the algorithm, the more those tricks break because the easier they are to see through you can see this was not written by an expert, or this was written hastily or this was copy pasted together.

Or this was written by somebody for whom this is not even their first language.

And so They don’t speak natively and there are at least telltale traces in the content that yeah, this is not this is not look or sound like authoritative content written by an expert with deep subject matter knowledge on the the question being asked to it.

And you don’t need a ton of content to be able to reflect that right? You don’t need to write 10,000 words to reflect that.

There will be words and phrases and prepositions and grammatical structures that indicate you know what you’re talking about? Very, very quickly in your writing.

I compare what your 21 year old intern is going to write about the topic of say, a cardiac devices versus what your chief surgeon is going to write and say, Hey, I need to progress me to you on the latest trends in in cardiac devices.

You’re going to get fundamentally very different language even in 300 400 500 words, right?That’s about two minutes of somebody speaking, ask your intern to speak for two minutes about cardiac devices, ask your chief surgeon to speak, you’re going to get really different content out of their mouths.

So BERT changes the game for low performing marketers.

Because what it does is effectively, it makes the great state great, but it makes the media or really become worse, right in terms of search performance.

It also changes the game for a lot of low performing marketers because it’s harder to gain right? When you have algorithms that can do reading at the sentence, paragraph and document level.

It is harder to just add in a couple more keywords.

Hey, put some more keywords in there.

In the old days, you know the old joke, an SEO expert walks into a bar Tavern pub, and so on and so forth.

Those jokes existed for a reason, because that’s what you need to do.

That breaks down when you’re looking at the document level that breaks down when you’re looking at the paragraph level.

If it doesn’t read the way, the authorities in the space, right? It will not do well.

It’s harder to game.

So Danny is 100%.

Correct? Nothing to optimize for BERT.

If your stuff is great if your stuff is not great, there’s a lot to optimize for.

There’s a lot to improve into images.

Question Is anyone testing? Yes, we did some testing yesterday on a page that have lost over 2000 visits this year from organic search.

I ran the process outline in a post over TrustInsights.ai put a link below if you want to read the the 10 step process for this type of optimization.

And what we found was that there were some key topics and points that were not in the page that were in the competing pages.

topics that made logical sense but just weren’t in the original piece.

Maybe for clarity, maybe for narrative flow, whatever the case may be, what was in everything that rank well was not in this target piece.

So we made some changes.

Now we wait to see what’s happening.

And we wait to see how it goes.

I’m running some tests on my own.

The transcription software that I use has continued to improve.

I just did an A B test with a year over year transcript.

And what is coming out now reads much more naturally, than the way it was transcribing a year ago, I use otter.ai.

And so for me, I have some work to do to go back and clean up older transcripts, we optimize them to sound more natural.

So those are some things to think about.

Again, Danny’s right, great content doesn’t need to change.

But if your contents not great, it needs to change a lot.

And for most marketers, that’s a very, very, very long road to all So, great question.

important question.

There’s a lot to unpack.

There’s a lot to watch in this space.

And you should absolutely be playing with all these natural language processing tools so that you can better understand how they work and apply your learnings to your content.

As always, please subscribe to the YouTube channel the newsletter, I’ll talk to you soon take care want help solving your company’s data analytics and digital marketing problems.

This is Trust Insights.

AI today and listen to how we can help you

You might also enjoy:

Want to read more like this from Christopher Penn? Get updates here:

subscribe to my newsletter here

AI for Marketers Book
Take my Generative AI for Marketers course!

Analytics for Marketers Discussion Group
Join my Analytics for Marketers Slack Group!


Leave a Reply

Your email address will not be published. Required fields are marked *

Pin It on Pinterest

Share This