You Ask, I Answer: The Impact of AI on SEO?

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 It may be out of date and no longer relevant.

You Ask, I Answer: The Impact of AI on SEO?

Max asks, “Your opinion on SEO and the impact of AI in the future?”

You Ask, I Answer: The Impact of AI on SEO?

Watch this video on YouTube.

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.

Christopher Penn 0:13

In today’s episode, Max asks your opinion on SEO and the impact of AI in the future.

So this is a, this is a complicated question.

And it’s a complicated question because organizations like Google have said, Hey, we’re going to penalize content that is generated by a machine that doesn’t add any value.

So there are any number of services out there that will use natural language processing, and essentially take existing content and reprocess or remix it.

Some was really bad, like, really, really bad.

There’s this one bot that scrapes a number of popular blogs, and it attempts to rewrite those blogs, but it finds the most awkward synonyms.

And you can tell pretty easily that it’s machine generated, right? However, what makes this question complicated is a question of skill, let’s say a human right, it goes from, you know, face rolling on the keyboard to Pulitzer Prize, right? Those are sort of the the spectrum of writing machines right now are kind of out here, right? So here’s face rolling on the keyboard, here’s, you know, competent, but not great.

Google Webmasters guidelines actually has a expression for this nothing wrong with nothing special.

And then of course, appears appeal to surprise.

The challenge is this.

for search engines like Google, it’s easy to spot the stuff down here, right? It’s easy to spot the stuff that’s barely more than face rolling, that is clearly no value add, that machines are generated programmatically using Yes, some machine learning and AI, but the outputs not great.

The output is pretty inept, actually.

But every year, the bar of what machines can do goes a little bit higher every single year.

And we’re at a point now where machines can create credible, mediocre content, right, that is indistinguishable from human content.

When you look at it, it passes the Turing test, you can’t tell by reading it was this generated by a machine it was just generated by somebody who doesn’t like their job, right? You read this and go.

So and so is proud to announce another flexible, scalable, fully integrated turnkey solution for blah, blah, blah.

It’s the marketing copy that we all see that we all think is not great.

And so the challenge for search engines, in terms of the arms race of detecting these things, is going to reach a point.

Now, this is my opinion, this is my opinion only.

But I think it’s going to reach a point where computationally, it doesn’t make sense to keep trying to identify AI generated content.

Can you do it? Yes.

Is it worth the compute cycles to do it? No, not past a certain point.

And that certain point is, if a machine writes genuinely helpful, useful, original content that I as a human can’t tell the difference, I don’t know for sure, if a machine voted or human wrote it, then a search engine is going to have a real hard time determining that as well, particularly at scale.

One of the things that we forget a lot when it comes to machine learning and AI when it comes to big tech companies like Facebook and Google etcetera, is that they not only have to employ this technology, but they have to do so in a cost efficient manner in a computationally efficient manner.

And that means that the cutting edge techniques in many cases are too computationally expensive to do at scale.

Right? When you look at something like a T five transformer, or when you look at a model like GPT, three, or DaVinci, or any of these really fancy text models.

They don’t have the same computational constraints that someone like Google does, Google has to ingest billions of pages a day.

And to scam any more than a sample of them is computationally infeasible.

Right to develop extremely complex algorithms to detect and discern, did a human right this or did a machine write this when you consider useful content? Again, it doesn’t matter who wrote it.

Christopher Penn 5:01

doesn’t matter if it’s helpful or not.

And so Google is looking at with its most recent algorithm update, which is this has been recorded in September 2022.

The helpful content update, there’s definitely some content out there is machine generated, that does not help anybody, it is just garbage.

And that’s easy for a search engine to spot it’s easy for you and I to spot where we run into trouble is when we’re not sure anymore, like, so what happened here did a machine right, this did human right, this, it’s not bad.

And because of that computational disparity between what Google has to process at scale, and what an AI model that’s very sophisticated, can process on its own and not have the same scale constraints, the AI model is going to win, eventually, the quality gets better, so good that Google will not be able to keep up, they may not already be able to keep up with the best stuff.

For example, I can download and run any of the you Luthra AI language generation models and run them ran on my laptop, or run them on Google colab or run them anywhere.

And they can generate, you know, couple 100 pages of text really quickly.

Now, it may take an hour or two for my machine to crank out that much.

But that’s okay, I can wait right? I can wait for 200 pages of okay text.

But the quality of that output is going to be better than what Google can look for at scale.

So what should you take away from this? The AI writing tools right now are still not great.

They can produce really good, mediocre quality content that can produce mediocre content that you couldn’t tell if a junior staffer wrote it? Or a machine wrote, right? It’s, it’s that good that it’s just average, right? And most of the content in the world is average, most of the content of the world is mediocre.

Read press releases, read corporate blog posts, read thought leadership blogs.

I mean, it’s the same old stuff, in a lot of cases, be customer focused, right? We’ve been saying that for what 80 years, be customer focused.

Can a machine write that article as well as the CEO of a Fortune 50 company? Absolutely.

Because you’re not gonna say anything new.

So the challenge for you as a marketer, for me as a marketer is not only to create good content that’s above mediocre, but to create original stuff, stuff that is truly unique stuff that is truly has not been seen before, and is not a retread that doesn’t add value, right? The world doesn’t need another blog post on being customer centric.

The world doesn’t need another blog post on being more human and social media, the world doesn’t need, you name the marketing trope of your choice.

And there is a risk that if you’re just cranking out the same old swill, you might actually get flagged by the helpful content update as being machine written, like if what you’re writing is, so copy and paste.

So templated, you won’t actually be detected as a bot when you’re not.

So you’ve got to up your content quality machines will continue to improve what’s happening right now with transformers and diffusion models in AI, is game changing, machines are creating better and better content every day.

And for those of us who are creators, we’ve got to keep upping our skills, we’ve got to keep becoming better at our craft to stay ahead of the machines if we don’t have the machines going to do our job, or good chunks of our job.

And we won’t, right.

And I’ve as I’ve been saying for a while, an AI isn’t gonna take your entire job, it’s just got to take like 60% of it.

But if there’s 10 of you at a company, the company doesn’t need six year, right? Because you can take that 60% of labor that it’s machines doing and they can a company can say yeah, we can we can afford to downsize.

So machines won’t take your entire jump does take big chunks of it, but it will be enough that it will be a scale issue for you.

flipside, if you are a lean mean scrappy startup, you will be able to punch way above your weight with the assistance of machines right.

Christopher Penn 9:47

If you can have a machine generating ad creative, you know 1618 100 pieces of ad creative overnight and in using a diffusion model if you could have a machine writing a 150 200 Blog posts a day.

You know, again, we’re not talking about huge shall million piece datasets, we’re talking 100 pieces.

But if you’re a startup, and you’re a team of three or five or 10, you can with the assistance of machines that look like your team of 500 Behave like a team of 500.

So the onus is on us to scale up as individual creators, and the onus is on us to master the use of these machines so that we can scale ourselves our creativity, and add that final polished machines inevitably, struggle to make.

That’s the future, as I see it right now.

And that again, this is my opinion, this is my opinion, but that’s the way I see things going, where machines are going to create, they today they create the first draft.

They’re gonna evolve to create second third draft.

And yeah, depending on the content type, they may be doing final drafts in a couple of years.

So keep an eye on that.

Really good question.

We could spend a whole lot of time on that, but I think that’s a good place to stop for today.

Thanks for asking.

If you’d like this video, go ahead and hit that subscribe button.

Machine-Generated Transcript

Christopher Penn 0:13

Christopher Penn 5:01

Christopher Penn 9:47

Comments

Leave a Reply Cancel reply

Pin It on Pinterest