Something big is happening to the way people use Google and it’s not what anyone expected.
A new study just dropped that tracked how 133,000 real people used Google over a 19 month window and the data completely destroys what we thought we knew about search behavior.
Remember how everyone talks about long tail search?
Well, turns out nearly half of all Google searches go to just 10,000 queries.
The concentration at the top here is pretty insane.
But here’s the really crazy part.
A third of all Google searches aren’t even searches anymore.
People are just using Google as their internet browser.
In this episode of the Authority Hacker Podcast, we break down what this means for your SEO and marketing strategy.
Plus, we take a look at OpenAI’s mind blowing new Sora video model that can create custom video clips from text prompts and new data showing that one in 10 Americans are now using AI as their primary search engine.
The way people find information online is changing faster than ever.
So let’s dig into what that means for all of us.
So our first story is from Rand Fishkin, who’s done some new research analyzing 332 million queries.
Interestingly, over 21 month period, and he’s got this data from Datos, which is a company run by SEMrush, who have this anonymized clickstream data.
What are the main findings from this study, Gael?
The first thing that he went through was essentially the long tell.
It’s like, and I’m actually going to scroll through the article here because I want to go through what he’s done.
He analyzed all these queries.
He has a lot of data.
He even used AI.
That’s interesting.
But the first thing I want to talk about is actually the long tell study that he did, actually, which kind of like goes against conventional wisdom when it comes to SEO, because we all are told SEO in a way that like, you know, there’s a lot of queries in the short tell, but really long tell makes most of it.
And that’s why SEO makes sense, because you can dodge the competition of big size by targeting long tell keywords, get very highly qualified traffic, and still do well even though there’s high competition.
That’s why we’ve been told.
But I think based on his findings, this has changed quite a bit, because the numbers that he shows is that the top 148 queries make up 15% of the volume.
And you can see this graph here that I have on my screen here.
Let me make it bigger.
Well, essentially, these 148 queries make up 47.86 million searches.
So you can see he has these different buckets with how big they are.
And the long tell– These are queries like YouTube, Gmail, Amazon, weather, speed test, calculators, the stuff you would expect lots of people search for.
Yeah.
And it’s like a lot of it, as you can see, is like navigational queries, right?
It’s people just trying to go someplace and go there.
But what’s interesting is like we kind of have this vision that like long tell is very big.
But like what he’s trying to prove in his point is that essentially, it’s not as big as you think, because if you look at the bottom bucket here, which is keywords between 100 and 250 searches per month, which is interesting, by the way, because they didn’t go below that.
Like they didn’t have data for that.
And my expectation is the last bucket, if you had like 100 to zero, for example, it might be quite big, actually.
And so like that’s skewed a little bit.
Well, they did have that data for one month.
Like this was a privacy thing, I think, because if you give people– Because you could find who it is, yeah.
Yeah, because you might kind of incriminate yourself by the word you use in search.
So they were very strict about that.
Like he had to delete the data for, I think it was September 2024 afterward.
But he does have that data for one month, right?
Yeah, but still it’s like it’s interesting to see that essentially you need 153,000 keywords that still make up about half of the volume of the top 148 keywords.
So it’s like it shows you that while long tell is decent, and you can see it’s growing towards the end.
So you could imagine like the last bucket would be pretty high up here.
Like it’s still not what you’ve been taught maybe 10 years ago when you learned SEO.
And what really this shows us is that essentially these big platforms are winning.
And like that’s where people are now going to consume content.
And they essentially use Google like a net bar where they instead of typing the URL of the place where they want to go, they just type the name.
It takes them to the website.
It’s almost like Google is a DNS, right?
It just links the name to the IP of the server where they find what they want, right?
So yeah, this is kind of like confirming something that we observed when we prepared a lot of the news videos, for example, where we found that 90% of the time spent on mobile in the US was spent in dedicated apps.
And only 10% of the time was spent on browsers, which again is kind of like the same as direct searching into your browser on desktop.
So yeah, that means big platforms are winning.
And then small discovery searches are essentially less big than we think.
However, there’s actually been, and actually I didn’t open this, but let me just open this right here.
There’s been a video by a guy called Will Creslow, who now runs a company called Search Pilot, but brand a company called Distilled before.
We’re quite closely affiliated with Rand Fishkin in the past, right?
They’re friends.
They’re friends.
They’re very good friends.
They speak in conferences together, et cetera.
And like he looked at this, like I’m not going to play the volume, but he looked at this data and he kind of like remapped it, right?
He was like, if we look back, I’m just going to pause this and go back at this graph.
He’s like, look, the buckets are not the same because here you have like between 50 and 100,000 searches bundled together, right?
So that’s 50,000 volume span.
Whereas if you go here, for example, between 500 and 251, you only have 250 searches per month volume span.
So essentially this bucket here has way more volume, search volume span than this one, right?
And so what he did is he remapped it to essentially make similar size buckets.
That’s what he did.
And now I’m going to fast forward the video and I’m going to show you, like you see his graph now shows way more long tail and shows you a little bit more of that when you normalize the space between the buckets that he did.
So he kind of like counteracted runs point a little bit and it’s not completely untrue.
However, it’s still like, if you think about the top 150 keywords and even like the ones that come after that 178 in a second bucket, these still eat a significant amount of search volume and way more than we’ve been taught in traditional SEO.
So basically what it says is long tail is just not as powerful as it used to be, but Wilcresto kind of like moderated that and said, well, it’s still not that bad when we normalize it basically.
That’s what most people, for most people watching this, there’s very little they can do to get traffic for or influence the top 500 results.
Yeah, you can get that.
Yeah.
So it’s not even a game that most people are really playing in.
So the entire game is played sort of in the middle towards that long tail.
So does it change anything?
Well, it’s like essentially what this tells you is that more traffic is shifting towards going on platforms than it is towards going for like informational queries.
And there will be more about the rest of the study that actually confirms that.
So what I suggest is we go back to it and we actually keep going through that.
But yeah, you can see he has this graph as well of like distribution of 232 million Google searches over 21 months.
And you can see like a lot of like you see that this is the percentage of search volume.
And you can see 46% is on the top 10,000 queries.
So it’s like it’s still pretty skewed over these things.
Now let’s go to the next section, which I think is quite interesting.
Which is brand searches versus generic searches.
And so again, that’s an interesting thing because brand searches is when people are searching for a company, they know where they want to go.
They type this thing, they click on the website, and then they just go do whatever they wanted to do on that website.
It’s essentially just a time saving exercise instead of typing coca-cola.com or whatever.
Yeah, exactly.
It’s just that.
And generic searches are searches that do not mention a brand where people are essentially searching to find something new because they know what they want, but they don’t know who can help them achieve what they want.
That’s kind of the point.
And what Rand said is he thought two-thirds of searches would be brands, but he was actually a little bit over enthusiastic on brand searches.
But still, 44.19% of searches are brand searches, which is massive.
It’s like almost half of all searches.
And 55% are generic, which means that again, you can take all search volume that exists on Google, and roughly half of it is something you can’t really compete for because it’s a brand.
This brand is probably going to rank for this search, and there’s not much you can do apart from maybe ranking for maybe like VS queries, that kind of stuff in the search space.
It’s something you can’t game with SEO.
So that’s starting to paint a bit of a grim picture in terms of SEO because we have the first point, which shows you that essentially a lot of search volume is concentrated on these keywords you can’t really rank for because they’re also navigational.
We also have a stat that we didn’t mention so far, but 60% of searches now have no clicks.
So it’s like you can take that.
And on top of that, 44% of searches are also brands.
Now, these things don’t really add up.
They overlap, but not completely.
If you were making circles, they’d overlap two thirds on top of each other, and it’d still be a bit of ground.
It’s very hard to tell how much ground is left for essentially generic searches that have clicks and that essentially you can compete for, but my guess would be around 20%, 30%, something like that.
So basically, when you do SEO, you get to compete for one in three Google searches, more or less.
That’s where I would go, and that’s a very wide guesstimate.
There was also quite a wide distribution in terms of niche.
So he looked at arts and entertainment games and sciences taking up a significant percentage, 37% of topics.
And he classified these as low monetization, perhaps to the exception of if you’re a high volume site running ads or something.
Yeah, they actually had it here.
So you can see the volume per niche, basically.
You can see art and entertainment is really big, actually, like 17.5% pretty much.
Which is like– Is that gossip magazines?
Pretty much.
Like entertainment.
What’s Kim Kardashian up to today type thing?
Yeah, it’s pretty much stuff like that.
But the thing as well is if you think of these kind of queries, they are heavily templatized at this point.
If you go on Google and you type like “Aina Grande SNL” or something, that’s something that happened recently.
I’m on perplexity, sorry.
“Aina Grande SNL.”
It’s an interesting point.
How many people are moving their default search engine away from Google these days?
But you can see, I typed this.
And I’m not even in the US, right?
It’s like I don’t even have AI overviews here.
But you can see essentially, this is the kind of query that Rand is talking about, which is a bunch of images, some kind of trailers and clipsing that takes me away from Google search.
A YouTube video, like one article from a big newspaper.
Okay, I have Ahrefs on the right, but then I have a block of YouTube videos.
And then it’s like YouTube, YouTube, Facebook.
People also ask one magazine, two magazines, TikTok, and that’s it.
So there’s literally like two or three links to websites.
So this has really become becoming more of a portal rather than 10 blue search results, right?
That’s the thing, right?
It’s like with what’s left, Google is essentially making these new layouts that don’t really lead to too many websites.
Now, I’m not saying there’s no space left to actually get search traffic or anything like that, but you can see essentially the main places where traffic is going these days is not to a place where websites are prominently displayed and people will click through to a website.
They will probably either click to a platform or they click to the brand that they’ve been searching for.
Essentially, people are finding information on YouTube, on social, they hear about stuff on like a WhatsApp discussion, whatever.
Then they go on Google, they type the name of whatever they search for, they click through it, and then they just have, yeah, have this kind of like portal with like a bunch of stuff.
Like when you Google a brand, it’s the same thing.
It’s kind of like almost more interesting than going for the brand itself.
Like if I type Adidas or something, if I know how to type Adidas, I don’t remember this one, D or 2D.
Like I get more information than just Adidas itself.
I just get like where the shops are, for example.
I would get reviews, I would get all of that.
It’s not even a good query, but like you get the idea.
You get a lot.
It’s a portal, basically. – The celebrity one’s interesting because it’s like they are a brand themselves, right?
You don’t discover celebrities through Google.
You hear about them through the media, through social media in other places as well.
Ran’s got this little graphic at the end, which is his theory of web journeys.
And there’s four stages to consume, which is you’re consuming content on social media, YouTube, podcasts, news sites, email newsletters, webinars.
And then once you kind of are already familiar with something, could be a person, could be a topic, could be a brand, could also just be a type of product or a concept, right?
Then you go and search for things.
And that’s very interesting to me that search is second, because we’ve always thought of search as being the kind of like the first step in trying to find something, right?
But it’s actually the path starts much earlier now.
And we’re seeing AI tools and LLMs kind of get in before the search as well.
We talked about that last week in the show as well. – Yeah.
And then there’s the compare phase essentially where he says like people go to Reddit, go to like, they go back to social quite often actually.
Like they might check videos and so on.
And eventually they buy.
But the thing is with search as well, what’s interesting is that I think it may become a bit over credited for conversion stuff.
So people would be like, oh, like, okay, like you’re doing social is cool and stuff.
But like, if I look at my conversion data in Google Analytics or whatever platform I’m using, I’m still seeing lots of conversions for search.
But like if a lot of that search traffic is people seeing something somewhere else, then Google going to Google to essentially replace the address bar, click on the Google link. – It’s cannibalizing your direct traffic. – Like the decision was made from the content they’ve seen on another platform, not from the last click they’ve made on Google.
And so it’s like, it might actually start, it might be over reporting on conversion for a lot of companies.
I think that’s going to excuse people’s effort towards SEO going forward, because they’ll still see conversion data coming from Google, even if traffic’s going down, whatever, mostly because of that and the other channels will not be credited.
So yeah. – It’s almost the same as, you know how a lot of companies will bid on their own brand as like a paid keyword.
And then suddenly their Google ads start reporting great conversions, but they’re just stealing their own traffic from organic.
It’s like, well, organic stealing it from direct.
And it’s just like, you know, passing the buck. – Especially as social media platforms do not allow for very much links to things.
It’s like, you know, TikTok, Instagram, et cetera, like you were watching Reels, whatever, like it’s not like you can place an obvious thing people can click on.
So I guess their default is just open the browser, search for the thing, click there. – Yeah. – And then the conversion data, you know, comes from that.
So it’s like, it’s interesting. – I’ve seen a lot of brands these days on their YouTube channels, instead of telling you go to carwow.com, they’ll say Google CarWow or Help Me CarWow or something like that.
And so they’re essentially creating all this extra branded traffic in that way as well.
And what’s also interesting is that one of the takeaways or insights from the helpful content updates Google released last year was around branded traffic.
And I think there’s maybe a case to say that SEOs have misunderstood the impact of that or what that actually means, because it was never really branded traffic.
It was actually navigational traffic, right? – Yeah.
You know, they gave the example with the business card.
They were like, oh, just distribute some business cards and people will search for you.
And people were just making fun of Google.
Like, oh, business cards ranking factor, et cetera, right?
It’s like, it was just a joke.
But really, I think what they mean is, I think they’ve understood that essentially, the consumer journey has changed and a lot of Google is now navigational.
It’s not really as much of a discovery tool as it used to be.
And so if they identified that, then it’s very much something they’d be looking for to reward websites with search traffic, because that just tells them that people are, want these brands, they want something about them, et cetera.
And that’s why that could be a reason why content sites tanked, right?
They just identified that, hey, people use Google in a way to just find the stuff they’ve seen elsewhere.
So we should reward these sites because obviously the more people search for these things, the more people want it.
We have essentially people’s moderation by they just vet the offer, et cetera, and they just go for it.
So this is a very high quality signal, except content sites that were just made to rank on Google, like you would forget about them five minutes later, you couldn’t even tell what their name is.
And then resulting in like having very low performance against these metrics resulting in tanking.
And so it’s possible that’s what happened.
And then Google was like, had no intention to tank these websites, but by upping these new factors that these sites performed poorly against just destroyed them unintentionally.
And then you have Danny Sullivan just apologizing on Twitter for two years after that. – It’s also kind of like this idea of people trying to do SEO, not doing so well in SEO, and those people not doing SEO and doing all the other stuff, being successful with it. – Yep, yeah.
You see a lot of creators like Lucy that I interviewed on the podcast, right?
Her site is not very good, but she’s very big on socials.
She has a lot of branded search traffic and her traffic has been going up consistently ever since these updates. – And again, she’s really big on YouTube, right?
So it’s hard to get links off of there, people watching in different devices, things like that.
So people will just Google for her and therefore has lots of branded traffic. – Yep.
And so, but like that, if this is how Google works, it’s fundamentally changes how you should approach SEO.
SEO now becomes this kind of like technical thing, like make sure your site is scrollable, make sure like you deal with all that stuff and like your pages are like essentially high value so they can rank higher.
But a lot of the stuff that would make you rank is almost out of the scope of the traditional SEO role. – So are you saying only doing SEO as a kind of marketing strategy in isolation is no longer a thing or shouldn’t be a thing? – I mean, we know that it’s less and less of a thing, right?
It’s like people who like go hardcore on SEO and do nothing else.
They haven’t done well in the last two years.
Like let’s just not deny that.
It’s like everyone’s down that has been focused on 100% SEO.
Now it doesn’t mean, but the thing is like if you solve these other things, that there’s maybe a hope that you can actually get more Google traffic.
But by that time, Google traffic should be less meaningful to your business as well.
Like because it’s user metrics, it’s like user engagement that would determine your success.
So you need to have created something that people engage with and it’s much easier to monetize when people actually care about your stuff to the point where they would Google it, right?
So it’s like, basically the solution to your Google traffic problems is to actually not care about Google traffic and eventually it’s actually gonna come back and be something that boosts your business.
But it’s still– – It’s really hard to care about something but not care about something to get the result.
That’s like, yeah, I can’t see too many people being successful with that. – Yeah, but I think the fundamental thing that we learned from this study is Google is now very much a navigational engine, not as much a discovery engine.
There’s only a small percentage of queries that actually lead to websites with articles, the old school way.
It’s probably not the way to go forward.
And if you want to do well on Google, you probably want to essentially emulate this way of finding you because there’s a good chance Google measures that and helps and gives you more traffic when you’re doing well with this. – And now for a quick word from this episode’s sponsor, digital.pr.
They’ve just launched the world’s first subscription-based digital PR service that makes premium 100% white-hattling building accessible to anyone through a mix of reactive PR, expert commentary, and data-driven campaigns.
They guarantee a minimum of five to 20 high-quality links from top-tier publications every quarter.
And this is not some shot-in-the-dark approach.
It’s backed by Search Intelligence’s proven track record and is made possible by the world’s largest digital PR team, now accessible at the click of a button.
Unlike traditional agencies that require huge retainers, digital.pr offers transparent monthly plans starting now from just 700 pounds per month.
That means no long-term commitments and no hidden fees, just guaranteed results.
So if you want to level up your link building without breaking the bank, head over to digital.pr and join the revolution that’s making premium digital PR accessible to everyone.
And now back to the episode.
So, I mean, that’s pretty much this for this story.
Anything else you want to add? – No, let’s move on to our next story, which is OpenAI has released its Sora model, which is its text-to-video generation model.
And it’s pretty wild what it can do.
I want to temper things a little bit though.
I think there was maybe a bit of an expectation that you would be able to put in a two-line prompt and get a Hollywood movie out of it, but it’s absolutely not that at all.
Currently, there are some limitations.
You can generate clips up to 20 seconds long.
And even doing that can take several minutes to process.
If you can even get in, at the time of recording, we haven’t been able to get access to it.
It’s completely overloaded.
I mean, shock horror.
A lot of people wanted to use this thing.
Never saw that one coming, but apparently Sam Altman was surprised at the demand.
Or maybe it’s just a marketing play.
So more people want to get in on it.
I don’t know.
But my first reaction to this is it looks pretty good.
There are some kind of oddities that don’t look all that natural sometimes.
But if you are a B-roll company providing B-roll content, I think you got to be pretty worried about that right now.
Because this is essentially anyone can create any kind of cool B-roll for a video anytime they want easily.
Yeah.
I mean, okay.
Some stuff there.
So first of all, Sora is good, but there are models just as good that are available right now.
There’s like Chinese models.
There’s one called Minimax, for example, that actually does videos that are arguably as good.
There are even like open source models that are not very far from it.
So it was amazing when they revealed it like nine months ago.
They revealed it.
And now they finally release it.
What makes Sora good is actually the suite of tools that they give you with it.
So you can remix videos.
You can go from image to video.
It’s just easy to use.
The 20 seconds clip up to 1080p.
It’s only on the Chagipiti Pro account.
So that’s the new one that costs 200 bucks per month.
If you’re slowly starting to see that account price inflation, we’ll have the Entrifrise one come out soon.
That can do a minute, I guess.
And if you have the 20 bucks a month one that most people have, if they pay for Chagipiti, you can only make a five second video up to 720p, actually.
So it’s like, and you can only make 50 generations per month.
So really, if you’re gonna be using this professionally, then you will have to pay for it.
You do get to remove the watermark as well, which I saw was very funny.
If you have the $200 per month account, which it’s very funny because like the point of the watermark is essentially AI safety preventing misinformation and all of that.
But for 200 bucks per month, we will give a shit about that. – It just so happens to align with their monetary incentives as well.
It’s funny how that works. – So like safety has a price apparently when it comes to misinformation. – I think what is cool though is the, if you go to Sora.com, they have a page with all the videos which people have been creating and you can go in and you can look at the prompts which people have used to create those.
So it can be a good way to kind of learn from others on how to prompt this to get cool looking videos and kind of learn that way. – Yeah, I think, I mean, overall, it’s gonna be a good tool.
Like my prediction is this is going to make Facebook even wilder than it has been today.
Because I mean, if you go on Facebook these days, AI images have been dominating.
A lot of people essentially run the Facebook creator program where they get paid for impressions.
You don’t even need to drive traffic to your site.
It’s just like basically generate engagement on Facebook with your content and you get paid very little amounts but it’s very easy to get lots of rich on Facebook.
And so people have been using AI images to do that.
We’ve made a video on HNews about this.
And my speculation is that people are gonna buy these $200 per month accounts so you can generate videos without a watermark and you’ll be able to generate crazy engagement by making crazy videos the same way images have been going around.
And that’s going to be like kind of like a side hustle for a lot of people. – Aside from spamming Facebook with these videos, are there any other real world use cases you think for business owners or for marketers? – Spamming Pinterest is gonna be one of them as well ’cause Pinterest is video now.
I mean, I think it’s gonna be mostly Bureau.
Like when we do YouTube videos, we can make kind of like crazy comparisons, whatever.
Like, oh, it’s like being a mammoth in a porcelain shop or just invent something crazy and you can generate the Bureau as you’re saying it.
And it’s like, if it’s like a five second, six second Bureau, even if it’s not perfect in physics or something, it will illustrate content better.
And I imagine that this is going to have an API and that’s going to be built in directly into your video editing tools.
So like you’ll be in Final Cut or you’ll be in DaVinci Resolve or Premiere Pro and you’ll be able to prompt and then just generate a clip and just drag and drop it onto your timeline and illustrate what you’ve been doing.
So yeah, anything that you’ve been using generic Bureau for, you could potentially use Sora for that.
And 200 bucks per month for that is actually not crazy.
If you think about it, like people do pay more than that for Bureau. – You mentioned the physics and some sort of technical limitations there. – Yeah. – There are, you know, if you look at people’s hands and legs in the video, sometimes it just feels a little bit off.
Like it doesn’t look supernatural, it gets a little bit confused. – I mean, look at my screen right now.
Look at the video here with the smoke.
You see the smoke, like there’s basically some kind of smoke inside the glass, but look in the middle now, the smoke actually leaves the glass on the left.
Like see here now, like actually, like it’s actually broken. – So these kind of like collision mechanics or whatever you call it, it hasn’t really figured that out yet.
But, and they even said this in the presentation when they announced this is like, this is the worst it’s ever gonna be, right?
So it’s only going to get better.
I think if you compare it to the early days of AI images, when everyone was like, oh, look, that person has six fingers, this is terrible, but they quickly iron these things out and it will get better quite soon, I think. – Yeah, I agree.
I mean, like now hands are not a problem and so on.
Like, and it’s video is still like quite new.
Like we’re kind of like starting to hit a plateau on like text, for example, like it’s getting harder.
We’ve trained AI on literally all of the content humanity has ever created at this point.
But like for video, we’re kind of at the beginning.
Like there’s a lot of margin for progress and very likely two years from now, you’ll be able to make like footage that’s very hard to distinguish from real life.
And there’s no way like Shutterstock is not in big, big, big, big trouble right now, for example. – Yeah, yeah.
And you know, Shutterstock and the other stock footage providers, they were disruptors themselves for videographers who would have to go out and film this footage or take those photos themselves.
So it’s kind of like the next in the chain of disruption.
One thing though, that I think was particularly interesting is this is not available in Europe right now because I presume like GDPR, like data protection laws, something like that.
They didn’t get into the specifics of it, but do you think there’s a chance that here in Europe where we’re gonna fall behind because we can’t get access to this stuff? – As long as we have VPNs, we’ll be fine.
But yeah, it’s a problem.
Like actually Sam Atman tweeted that it’s annoying and there are features that they might even never be able to release in Europe due to regulation.
So it’s going to be a problem because essentially like if AI boosts productivity, then it means European workers could become less productive.
Imagine like hiring a video editing company like in Philippines who has access to these versus Europe.
Like they become even less competitive because they don’t have access to these tools.
And so on stock footage gets more expensive, whatever.
And so like, yeah, we’ll see how this goes, but it has been a problem for every company.
Like Apple Intelligence on iPhone has had the same issue where it hasn’t been released forward.
I don’t think it’s officially released in Europe until March or something.
I have it ’cause I kind of like hacked my way around it.
But for most people, it’s gonna be an issue. – I did have a look at the keyword best VPN for Sora.
And there were no articles specifically targeting that query.
So if you run a VPN site, get that article up today.
And I think you’ll make quite a bit of money on it because there’s gotta be a lot of people searching for that right now. – Yeah, anyway, I think it’s gonna be interesting.
I don’t think there’s like, I mean, you can put some stuff on your socials, et cetera, but like the truth is like fully AI generated socials are like not extremely engaging at this point.
Like if you are thinking like authenticity, but for Bureau for video is gonna be amazing actually.
And then for people spamming Facebook, I think that’s pretty much it. – Yeah, so if we fast forward two, three years, where do you think we’ll be with this then?
I mean, this must take a lot of processing power and presumably that’s why they’re charging a lot to edit or to create just basic short videos. – I mean, think about like Cintigia, right?
Cintigia is like really good at making kind of like an AI clone of yourself, but still a bit robotic in the way it talks and so on, but it’s still pretty good. – Gets the accents a bit off sometime, yeah. – Especially for me.
But the thing is, imagine mixing these two, right?
Being able to prompt the environment, but also use a clone of yourself.
Like you can make videos without, like you make a Google Doc almost, where you would write the script, you would write everything, but like instead of us being sitting in our room in a bit of a boring context, we can make videos of us like walking into in the middle of the streets of Tokyo as we present things and things like that.
And that’s going to be easier to do than it is today where we need to bring cameras, we need to edit it, et cetera.
You’ll be able to generate all that stuff with just text, which to me tells me that I don’t have much faith in text being like a very big medium for information going forward as video becomes easier.
And as people have shown they prefer consuming video than to reading text, basically.
So to me, that’s like another nail in the coffin of like blogging, et cetera, because these things are going to become so much easier to create going forward. – And we’re already kind of there with audio.
You’ve got like Google’s notebook, LLM.
I mean, it’s been around for a while now, but it’s very good at creates this like two person podcast on any article or any video that you want to watch.
So it’s essentially a way to kind of like distill that information.
I think we’re moving into a stage now where it’s like the underlying information and content is still important, but like the medium of how it’s presented can be modified like very easily almost at will.
So people can choose their preference for the way they want to consume something. – But you know what that means as well.
It means that like, while it’s still fairly easy to make it on like YouTube, et cetera, today, if you’re making okay videos, it’s going to be very challenging going forward when it becomes extremely easy to do these things.
Of course, people will be like, “Oh, here’s how to automate everything with Sora, like make a A-roll, B-roll, whatever, full automated in five minutes using make.com.”
And while these things will not necessarily get a lot of engagement if they’re like not worked on properly, it’s still going to eat into the market share of people who make legitimate stuff the same way it has happened with blogging and text all the time.
Like even if they slowly kind of like eat at the market share, like it does have a profound impact given enough time.
And I think that if you want to build an online presence when it comes to like video, social, et cetera, now is the time to get into it because when it gets very easy, it’s also going to be very easy to create the content.
It’s going to be very hard to grow your reach because of the overflow of content.
So yeah, build your profiles now.
That’s pretty much what I mean. – Yeah, everything’s about to change.
And it already is. – In a few years, yeah, for sure. – So this next story is from SEMrush, and they’ve teamed up with Statista and they’ve created this joint report about online search trends and the impact of generative AI on search basically.
The headline finding or one of the headline findings is that one in 10, so 10% of US users are now using AI first before they go to Google, before they search for anything.
AI is the first point of call when doing something.
A surprise at all by that?
Is that higher or lower than expected? – I think it’s higher than I expected, to be honest.
Like, you know, when I talk to about AI to people around me, they don’t give a shit usually.
So it’s like finding that one in 10 is going to be, is already using AI is quite a lot, but what’s interesting is kind of the projection by 2027, which would be one in four that would be using AI as their first point of research, which is, it’s pretty high.
And what’s interesting about this stat is, I think this is considering the state of things today, right?
It’s not considering any kind of advancement that’s going to come.
So for example, we know in the first quarter of next year, like the search function of charge should be free to everyone.
So it becomes like a real Google competitor.
It’s not just a paid thing when you need to pay $20 per month, which is a huge barrier, by the way.
We also know that there’s the next version of iOS that’s coming out probably end of this year, where Child GPT is going to be built into Siri, for example.
And so you’ll be able to talk to Siri, and it will just search, like it will pop up to ask if you want to search on Child GPT, and you click yes, and it will just give you the answer directly in the system, et cetera.
So, and it’s also not counting that Google might have to sell Chrome and not be the default search engine in there.
So it’s like, this number is growing, not counting any of these things.
I don’t think people have considered that when they answered the survey or when they prepared the survey.
So yeah, I expect it might be more actually.
Talking about our own use, and you sort of alluded to this earlier when you accidentally did a perplexity search, but you’ve changed your default search engine to perplexity, right?
Why’d you do that?
Because I get answers faster.
So like the thing is like my setup, like you can see my screen now I’m on the SEMrush tab.
So like, let’s say if I search, let me actually go on my browser.
Let’s say I search like SEO, then it goes directly to perplexity.
What I like about with perplexity is it gives you an answer instantly.
Like you don’t have to wait for the LLM result.
You actually get a link for navigational searches, which like Chagipity may not do.
And you get like a lot of links to click on.
I click quite often on perplexity.
I do go to other websites, and you just read this if you want the information.
But what I’ve done is I’ve also, if I make a new tab, I can type G, then SEO, if I SEO, and that’s going to actually search for Google for me.
So I’ve made like this little shortcut because I still use Google quite a bit.
And I know I want to use both.
And I think that’s kind of like what this report is saying as well, is that a lot of people are now kind of like using multiple platforms to find information.
But the way I would go is I would probably search AI first.
And if AI is unsatisfactory for information, then I may turn to Google search to essentially go and read human content if the sources in perplexity or Chagipity were not good enough.
I actually, I did the same, but I went back.
So I was getting frustrated because I found myself using Google for so many of its expected featured snippet searches, like looking for a stock price or weather or these things.
And this perplexity still does give you that relatively quickly.
But it’s just three or four seconds slower now.
I was getting frustrated.
So I set up, I did it the other way.
I set P space, we’ll search perplexity, and then I just do it that way instead.
Yeah, but it just shows that like essentially before you would go for Google for anything.
And now there’s a bit of like a decision point where it’s like, oh, what’s the best tool to use this?
I also have Chagipity, the desktop app.
And if I press like option space on my keyboard, it actually starts at Chagipity.
Honestly, I use that more than I use Google now.
Yeah, and so like, I kind of had this decision point where I’m like, oh, do I open a new type and search?
Do I type G and search or do I do option tab and search when I’m looking for something?
And like this wasn’t a conversation I had with myself, even when Chagipity came out, right?
When Chagipity came out, like no way I would have replaced Google with that.
It was just, it was hallucinating too much.
The answers were not very good, et cetera.
And so like, yeah, that considering these projections and considering that they don’t think about things like how the default may change in browsers, in operating systems, and just the progress of the product in general that might get better.
Like what if Perplexity starts caching some answers so you don’t wait for the answer, for example, like you might consider killing some Google searches in favor of Perplexity, for example.
So there’s much more margin for progress.
And what’s interesting is Sandar Pichai, the CEO of Google, I think in reaction to this, because I think they’re seeing the trend, said that search in 2025 will change more than search has ever changed.
Meaning you’re basically getting this in Google.
That’s what he’s telling me.
And they’re just going to run after this.
And that probably a lot less clicks will come from Google as a result, because these AI search results are just less, driving less clicks than traditional search historically, at least. – Yeah.
Now the SEMrush study also has some interesting insights into who the early adopters were that were using AI.
And unsurprisingly, the US had a pretty large percentage.
It was 25.6% of global users.
Young men in the 25 to 36 age range, which unfortunately, and sadly, we are no longer in. – I was seeing that, yeah. – The next thing is that these are the kind of moments in life when you start to realize time is no longer standing still, unfortunately.
But overall, 22% of adults were interested in AI. 28% were indifferent to AI.
And 60%, this is not a total amount, but there’s some overlap here. 60% said they would use it more if it were more secure. – Yeah.
You know what this reminds me of?
This reminds me of social media, actually.
So when social media came out, people were like, “Oh, this is not really trustable.
I don’t know what to do with it,” etc.
You should not use it, etc.
And now these are the same people who are glued to Candy Crush on their phone and just do the same stuff.
And it’s like, this has shifted significantly.
I’m thinking about my parents, for example.
I mean, they don’t play Candy Crush, but they use Facebook way too much.
And I see how the perception is probably going to change the same way.
And fast forward 10 years, it’s going to be the main way people find stuff, even though they’re hesitant at the beginning.
Yeah, I see. – This happens universally across with new technology, right?
Google Maps was a perfect example of this.
In the beginning, it got things wrong.
It took you down into farmers’ fields where bridges didn’t exist.
It was bad.
But eventually, people began to trust it.
And now, no one even knows how to use it.
No one has a map in the park. – I don’t know how to drive without it, honestly.
I could not.
I would not turn the car on if I didn’t have Google Maps.
Even in my neighborhood, that’s how bad I know things.
So yeah, I get it.
And so that’s the same thing that’s happening.
And that’s the thing.
It’s like adoption is kind of like an S-curve, right?
It’s kind of like slow at the beginning with early adopters.
So that’s kind of like the phase we’re in right now with these AI tools.
And then kind of like everyone starts using it.
And because people start using it, you see other people using it.
So you adopt it yourself.
That’s kind of like the super fast growth phase.
And eventually, you get to the maturity phase where it kind of like most people are using it.
But some people don’t want to change their ways.
The same way, my grandparents still listen to the radio and refuse to watch TV, for example, and stuff like that.
And then essentially, it’s very hard to get the last 15% of the market or something.
But I would not be surprised if even if we’re told in terms of progressive models, we would see this kind of high growth phase of the S-curve happen in the next two, three years when it comes to search.
Now, is Google going to be that tool because they change fast enough to adapt to people’s demand?
Or is another player going to capture this demand?
This one we can’t really tell.
What was also interesting about this is we actually did our own survey on AI usage almost two years ago now.
And the interesting thing about that is the survey respondents were mostly marketers and business owners, people doing SEO, people doing online marketing.
And they were essentially the most early adopter of the early adopters cohort, right?
So 69% of them at the time said they were already using AI in their business.
And this was January 2023.
So this was a month and a half after chat GPT came out.
And it’s crazy to see that not even 69% of people two years later are using, not even close to 69% of people are using it two years later in the general population.
So if you are watching this podcast, you are probably very, very, very far ahead of most people in terms of your willingness to use AI.
So that’s something to bear in mind as well about how much more change there still is to come.
Yeah.
I mean, and it’s like, there’s a learning curve to this, right?
It’s like the way I prompted charge GPT two years ago when it came out versus the way we actually use AI tools now is like all this crazy documentation we upload to them and so on.
Is that you’re so ahead if you already, if you’ve started this already, the point is like keeping up is quite challenging.
Like I spend a significant amount of my time just testing stuff and just to be at the edge of this.
But yeah, there’s lots of opportunities.
One other thing that the study identified that the Samra study was there’s a shift towards what they called natural language in how people search.
So this is instead of trying to program it by typing in a keyword, as some people may be used to people were just chatting with it like they would another human or another another person and interacting with it that way.
Now, what I found interesting with this is my understanding is that this has been a trend for before AI, like for years, Gen Z don’t use keywords.
They talk to systems in natural language.
Yeah.
And it’s going to be making it very hard to create content based on like keyword research and things like that.
Like these tools are going to become so much fuzzier.
And in a way, that’s why kind of like when we saw the long tail study, like I said, like the last bracket was probably really big because probably there’s going to be tons and tons more unique queries as people input with natural language versus typing keywords.
Just by the fact that it’s longer.
Yeah.
But that’s the thing.
It’s also impossible to almost to make static content that would answer all these queries.
That’s why generated on the flag content is going probably going to be the way to go, at least for text, like for video was still a while away from that.
But yeah, there’s it’s going to change everything.
One thing that was interesting in that study for me as well is like, essentially, they show that traditional search is just not growing anymore.
Actually, even even YouTube, right?
Yeah, it’s shrank this year.
Yeah, YouTube and Google like 2022 to 2024, they’re both lower in monthly active user, not by a lot.
And they’re still winning, like really a lot.
Like that’s there’s no question that they’re still dominant.
But like clearly, this is a mature space at this point.
Like traditional search is just not something that can gain like everyone who was going to access Google and YouTube has done it at this point.
And there’s very little market to capture anymore.
Whereas the AI search is massively growing.
And and I don’t see a world where Google just goes under like that.
But I can see a world where Google has like 50% market share.
And then the other 50% is split among two or three other players that that provide a different experience, basically.
And another thing that the study outputted was the OpenAI and Gemini have 78% of all the referral traffic that comes from AI tools, 78% comes from from these two.
Now, it doesn’t really talk about the total amount of traffic that all of these tools are sending in total and how that compares to referral traffic from from elsewhere.
So it’s like in isolation, it’s not that useful.
But something which I’ve seen come up a few times, and this was a question that got asked in our community last week, was around SEO services that are kind of are claimed to be optimizing for AI.
And I don’t know, to me, it feels a little bit kind of snake oily.
But there’s a trend right now where companies are trying to poach clients from other companies, from other agencies by saying, hey, is your agency optimizing for AI?
We can get you featured in chat GPT and all these LLMs.
And I’m just not sure that that’s even possible yet, right?
I don’t think there’s even like a method of optimizing for it that has been established and agreed on by most people in the– Or at least one that’s not fundamentally different to how you would do SEO in terms of technical.
I mean, the way I can see this work, probably I’m not going to say for sure, but it’s basically like they scrape the entire internet.
And when a solution is associated with a problem very often in the stuff they scrape based on the reputation, they essentially give that answer in the AI answer.
So the solution to ranking well on chat GPT would be to essentially be the consensus, the solution to a problem that people have and that most people would quote you.
And because they essentially repeat what they see, these chatbots, they would just repeat that when people ask the question about the problem, which I would be very– It’s a very broad problem to solve.
Like if you can solve that, like you’re making a lot of money.
SEO is not your problem.
It’s like SEO is really not your issue.
So it’s like I’d love to see how SEO companies claim they achieve that because I think it’s bullshit.
But it’s kind of like VCs, et cetera, just startups at this point to add anything AI to it so they can double the valuation.
I think something similar is happening in this space.
And some people kind of like started to say, oh, we’ll optimize semantically your article.
Like maybe you have synonym.
Like it’s basically LSI, right?
Stuff like that.
I just became like we optimized for AI basically as people started exaggerating more and more.
If you are a customer, I would say I would ask how they achieve this.
And if they can give you a convincing answer and potentially case studies, then I don’t think I would buy into that.
I would probably try to avoid this vendor.
That’s my thing.
All right.
Anything else you to add to the summer study?
No, not really.
I think that’s about it.
I mean, it’s a cool study.
I can see how this is going to creep up.
And we’re going to enter that fast growing phase of the S-curve.
But I can see Google staying number one for a very long time still.
Like both can happen basically.
All right.
So to wrap up, we’ve got a small story from Ahrefs here.
So they have added a new analytics feature.
What’s your take on this?
Like why might they be adding this?
I mean, first of all, I think it’s super smart, right?
It’s like Google Analytics 4 was released, like I think it’s going to be sooner year ago.
I mean, it was released much before that.
It was forced on people.
Mandated on everyone.
I think it was July last year.
Yeah.
Yeah.
Something like that.
But like everyone still hates it.
I still don’t know how to use it properly, honestly.
Like I’m pretty bad at it.
And it’s something I always tell myself I should learn and I never take the time to do it.
So a lot of people have been looking at a lot of alternatives.
I know like Clicky, for example, has been popular.
There’s one that’s called Pathomizing that’s quite popular as well.
Anyway, people don’t track their traffic as well as they used to because Google decided to kill the analytics platform everyone loved and put a shitty one instead.
And so I know– Just to be clear, like in its defense, right?
It’s not shitty.
You can’t do a lot more with it.
It’s just a lot harder to use and is perhaps missing some information when you look at it in certain ways.
But in terms of getting– It’s a note of price.
Yeah.
It’s like it’s not made for small guys, it’s made for the big guys.
And then you need to build custom dashboards that nobody has the patience to learn how to do.
Anyway, Ahrefs released that.
They understood the trend.
And they do that.
You need to install a new snippet on your site to get the data, which means they capture essentially more data about your website when you install this because they have JavaScript on your site and they can tell essentially everything users do on your site, which is why I think they also released this on a free plan.
Because as we said earlier, these tools, Ahrefs, SEMrush, et cetera, they need more data.
And data sources have been made more difficult to access through the rise of privacy policies and scandals, et cetera.
And it’s just not as easy as it used to be, which means they’re less accurate.
When you look at low-volume keywords, whether it’s 50 or 200, it honestly makes no difference in a tool like this because they’re just not that accurate to tell you that.
So Ahrefs essentially wants your data, so they’ve made you an analytics substitution inside Ahrefs.
And they’re good at making these kind of dashboards.
But the risk is that they become much more accurate for the keywords you’re ranked for.
And if you have some hidden keywords that nobody knows about that maybe have low search volume in Ahrefs, but you actually get much more traffic from, they’ll become more accurate on that.
They’ll report on that inside their tool to their paying customers.
And these people will be able to see your little secrets by using these tools.
So it’s like, it’s kind of a double-edged sword.
I’m curious to see how many people are gonna use this.
I think it’s a good move.
I think a lot of people will use it.
I would consider using it just know that you’re basically giving up your SEO secrets to them if you’re using the tool.
There’s always this argument that, well, are you just making your traffic data and all the data about your site more accurate for other people in Ahrefs who might be researching you?
Right?
That was at least the argument when the webmaster tools and wanting to get your GSE data came out there.
Now they have, to be fair, come back at the time and said that they’re not just transposing that one-to-one on there.
There’s a kind of blending process that goes on and it helps make all data more accurate.
And I believe them.
Yeah.
I actually believe them, but I think on a keyword level, they might just be using that data much more closely to what you’re giving them in terms of data than on the site level where it kind of like goes through traffic estimate algorithms and they probably keep that layer to keep it fair to people who use this.
Like for an obscure keyword, my guess is the volume is going to update much more accurately because it’s not necessarily tying back to your site directly.
And they kind of respect your privacy that way.
So it would be interesting to run an experiment.
I still think a lot of people should use it, especially if you’re not really tracking your data properly right now.
And if you, like, it’s only if you have like some keywords that are reported as very low traffic, but actually generate much more traffic to you, that’s why you’re putting yourself at risk.
If you’re not in that situation, I don’t see a downside to using this, especially if you’re not using GA4 properly.
That’s kind of my take on this.
And I expect they will improve it quite a bit because if they can become a viable alternative to Google Analytics and people use it, it’s going to be a lot of data for them and they need it for their product.
All right.
Should we wrap up there then?
Yeah.
That was actually our last podcast of 2024.
We’re taking a break for Christmas New Year and we’ll be back on Monday the 6th of January, where we’re going to be talking about what we think is going to happen in 2025 with online marketing.
So a bit of a predictions podcast.
I think people have a bit of an idea from the stuff we talked about today.
But yeah, I think there’s opportunities too.
I think we’ve been talking a lot about the things that have been falling apart in terms of search traffic, et cetera.
I know it’s not been fun, but I think there will be opportunities.
So we’ll talk about that in the next episode.
So thanks for listening, everyone.