#353 – 2025 Online Marketing Predictions

The way we use the internet is set to dramatically change this year.

And if you’re running an online business, you need to hear this.

Google just dropped a bombshell with its new Gemini model that has a context size big enough to handle an entire book.

Social media is being flooded with AI generated content that’s becoming harder and harder to distinguish from reality.

And the tools that we use to build and run our business, well, they’re getting supercharged with autonomous AI agents that can handle everything from research to customer support.

In this episode of the Authority Hacker podcast, we’re analyzing what other marketers have predicted will happen this year.

And we’ll share what we think is about to happen next.

We’ll reveal why major platforms like Reddit might be in trouble and what’s really happening with Google’s AI ambitions.

Plus how the rise of nano influencers could transform how brands reach new audiences.

Plus we’ll look back at our 2024 predictions to see what we got right and wrong.

So whether you’re running an agency, building a personal brand or growing an online business, this episode will help you navigate the massive changes just around the corner.

Welcome to the Authority Hacker Podcast.

And now your hosts, Gael Breton and Mark Webster.

Happy New Year and welcome back to the Authority Hacker Podcast.

In this episode, we’re going to be making some online marketing predictions for this year.

And I’ve got, as always, my co-host and co-founder, Gael.

How are you doing?

Yeah, pretty good.

Pretty excited for this year for all the new changes that we’re going to do in our business.

But I guess a lot of things are going to change.

One thing you didn’t mention by the way is that we’re also going to review some predictions from other people.

We thought it would be interesting.

We’re going to start doing that.

Then we’re going to present some of our own predictions.

And as a special treat, we’re also going to go back on the 2024 predictions, which we made last year and see if we got them right or not.

And there was a few doozies in there, so I’m looking forward to that.

Yeah, as usual, it’s probably going to be 50/50, but let’s see how this goes.

Let’s start right away.

Let’s just go into the first prediction from other people and talk about it.

Okay, so the first prediction we have is from Wes McDowell.

And his prediction is that custom GPTs will begin to be used more as lead magnets.

So this is really a commentary about the general rise of custom GPTs as content, as products, as things that people give away.

And there are a couple of specific examples he mentioned.

He’s got the case of like a financial advisor creating a GPT to give budgeting and investment advice.

Personal trainers offering workout plan GPTs and dog trainers providing behavior advice GPT.

So let me just get your thoughts on this whole segment.

Like are custom GPTs going to be a big thing this year?

No, I don’t think so.

And the reason why is because the point of the lead magnet is it feels like it brings immediate value with our effort on your end, right?

It’s like when we give a lead magnet away, like already very few people even download the lead magnet, you email them the link to the lead magnet after they gave their email.

And it’s like half the people never even click that link to get the lead magnet.

They just opted in for it.

That’s the amount of people.

That’s how lazy people are when I do these things.

And if you’re telling me that I’m going to get a custom GPT and I need to start prompting it to get the output, it’s already too much work as a lead magnet.

I don’t necessarily agree with that, right?

I think, okay, fair enough.

Not everyone’s going to take advantage of it.

But if you’re in some kind of person to person business, a dog trainer, a personal trainer, financial advisor, these are fundamentally different from what our business is or was.

I think that being able to speak to someone to get immediate help with your problem is helpful.

Well, you’re kind of getting, let’s say 80% of their perceived knowledge, right?

Straight away.

I would compare this, right?

We’ve both got dogs.

I have a vet app.

So if there’s ever a problem, I don’t have to bother going to the vet or it’s late at night.

I can book a call and in like 15 minutes, I can speak to someone.

This is, it’s kind of like this, but automated, okay, maybe quality is a little bit iffy or different because it’s not a real, real human.

It’s not just that, but like how do you take action and get answers fast?

Yeah, but how good is the vet going to be at making custom GPTs?

Like people suck at prompting.

It’s like I see people’s custom bots all the time.

And it sucks, man.

It’s like quite often the base model is better.

They put a lot of like junk inside the custom instructions and so on.

It just makes the bug behave worse than just default out of the box GPT, which is already trained on veterinary knowledge.

Like it’s like, if you’re unable to add like significant added value, it’s like plus the amount of defiance towards a from the general public, right?

It’s like, if I go to my mom and I’m like, here’s a custom GPT, she’s like, I don’t want this.

And it’s like, yeah, I think it’s like, it’s a very marketing thing in the marketing circles that sounds cool, etc.

Then go to the real consumers and I’m not sure they see the value.

So is it that people are not willing to engage with these types of products or the quality isn’t there for it to earn trust?

First, it feels like work because I have to go and ask the questions.

Whereas it’s like a good lead magnet.

Be like, hey, I’ve prepared your custom workout.

You have nothing to do.

Just go to the gym and do these things.

And that’s me, like it’s not an AI that came up with it.

It’s me personally as the real specialist that made this for you.

Just follow this.

This feels higher value to me as a lead magnet because I’m following that person or the AI of that person.

And then the second thing is like, yeah, it’s like, you need to be quite good at making these custom projects, custom GPTs so that they output something that feels significantly different from the base model they’re based on.

Like, you know, compared to GPT or compared to Claude, like, yeah, you need a fair amount of knowledge in there.

So it’s going to behave differently.

Let’s just strategize for a second there.

Let’s say you’re a personal trainer and you want to create a custom GPT.

Give us an idea of what that would involve.

That would involve like kind of like, I mean, first of all, you need to put all the core knowledge inside kind of like text files or something and well structured so that AI can reference to it when you’re adding asking stuff.

So for example, let’s say you’re talking about leg training, right?

Let’s say you’re the person that like, let’s kind of like two types of people.

They’re kind of the bodybuilders who are going to be very into like hack squats and machines that help you do that.

And then you have the power lifters that be like, no, actually, like you need to do a real front squat because that works your entire body.

And then it’s like you walk all these small muscles and that’s better for you, even though you maybe put less weight on it and your quads maybe not as big.

Like you need to start essentially just giving your vision of things inside custom files to that, to that, but so it starts reflecting on what you believe in and saying the things you would probably say.

And you basically build like a structure of fast.

So maybe you have something on like, you know, why compound movements are better than machines.

You’d have like a list of like a ranking of like your best leg movements, etc.

So that it can refer to that when it makes sensors, you make a bunch of these files and then you make your system prompt, essentially call out all the files that are provided to the custom GPT and tell it when to use it and how, so that it actually just structures and answer the same way you would probably do it.

So it’s kind of like building a software with multiple files that you call upon when you need it.

But it still has access to like the underlying information about basic biology and fitness in the LLM.

It does.

But if you let it refer to that too much, you’re essentially not really differentiating your custom GPT from just going to charge GPT and then the values.

What you’re talking about here is the personalization process.

Exactly.

It’s so important because otherwise like it’s like, yeah, I’m just going to charge GPT, has search, has everything, like why do I need your custom GPT?

You know?

So it’s like, I think it’s going to be like, it’s going to be used and some people will be excited by it.

I think it’s going to get old quickly as people executed poorly.

And then as a result, the opt-in rate is going to drop when you propose this kind of lead magnets because people will have had bad experiences.

And then eventually it’s just going to be like whatever.

I kind of disagree with it.

I think we’re going to see like a big wave of this where people are just throwing them left, right and center.

Yeah, there’ll be one or two that are good, but people will have bad experiences.

It’s kind of like when AI generate content first came out and everyone’s like, this is not good enough.

So they ignored it and then slowly got better and better and better.

You still need some sort of tweaking and like expertise to be able to get the most out of it.

But I think in the long term, this will be a big thing.

All right.

Let’s see how it goes.

I think it will be eventually, but I could almost see that like eventually, you have the memory in chat GPT, like you’ll talk to it a lot.

I will essentially build like your vision of things and it will be able to build a custom GPT for you based on all those discussions you ever had.

And that will make better versions than like the ones built by humans almost.

Like that’s probably not next year.

That’s probably like two, three years away at least.

Anyway, let’s jump on the next one.

Yeah, let’s move on to the next one.

This is by Matt Diggity.

And one of his predictions from his predictions, SEO predictions video was that Reddit was going to lose a ton of traffic.

He says that they’ve experienced massive traffic growth in 2023 due to…

Well, I mean, let’s talk about this.

Why did they grow so much?

Is this a blatant Google favoritism because of these kind of deals to share data and things like that for Google to train its AI?

Or is it just what users want?

First of all, I want to say that’s the easiest prediction to make because there’s literally no way out for Reddit anymore.

Yeah.

It’s like, I’m sorry, but it’s cool.

But this is the easiest prediction to make ever.

But fair enough.

I mean, why is it up?

Because people like it.

I think it’s like…

I don’t think it was intended by Google.

It sounds like it was.

But I think what happened is they just tuned up all the brand searches, et cetera, in terms of ranking sites up.

And then coincidentally, a lot of people were already adding Reddit at the end of their queries.

There was this article that went viral like a few years ago that’s like, oh, if you had Reddit at the end of your searches, you get much better Google results.

That plus just experience in general made that a lot of people typed queries plus Reddit.

It was always big.

Like even when VPN, et cetera, that was the thing.

That was a keyword with thousands of searches per month.

I’m not sure that Google has fully understood those people who are adding Reddit to the end of the search.

I think that while there are plenty of people that are looking for good, genuine, unbiased advice on products that they’re about to buy, they’ll add it.

I think there’s also a lot of people who find that the onsite search in Reddit just kind of sucks.

So they just go to Google and perform the search to find the thing they would have gone…

Google’s the navigation aid for them in this case.

Yeah.

It’s just because Reddit search has always been changed.

Apparently, they’re making an AI search for it eventually.

So who knows?

Maybe Google is going to lose a bunch of traffic to Reddit soon.

But yeah, it’s like, is it going down?

I don’t know.

What do you put instead?

There’s no way they’re putting a lot of content sites back.

In my head, content sites are going to live inside AI overviews where essentially they’re just going to quote them and you’re going to be on the right of the AI overview panel and you’ll be in the sources.

And then the rest of the queries, as we’ve seen even for our site, affiliate program queries, instead of showing the list that we used to make now they’re in the AI overview, and then they just show actual affiliate programs on these queries.

I don’t see them rolling that back.

It’s been a while.

It’s going further and further in that direction with every core update as well.

So what takes Reddit’s place?

We need to answer that question to know if it goes down.

Well, they seem to be kind of grouping Reddit, Quora, and they had this kind of discussion with forum…

Yeah, with Quora’s sheep. …feature, which was…

SERT feature, which was just Reddit and Quora basically.

Yeah.

Quora is going down.

You can’t really trust any of the other…

I mean, we can argue if you can even trust Reddit, but you definitely can’t trust some of the other kind of user-generated forums and stuff out there.

They just get like spammed way too much.

I think they might go down a bit, but I don’t think they’re going down a lot.

I just don’t see a platform that replaces them, unless, let’s say they make a deal with like meta and then start showing like Instagram posts.

So like, something that’s UGC, but that’s another platform.

It needs to be a big platform, right?

And the problem is like they don’t let each other crawl, each other.

So like Google cannot just like crawl off Instagram’s content and index it, which is why it rarely shows.

Or they always TikTok.

Let’s say TikTok is sold to an American company, let’s say to Amazon, and then they make a deal with Google to like expand the crawling ability of Google, and then a lot more TikTok content starts showing up in Google search.

Unless something like this happens, Reddit is still by far the best source for written UGC, which people seem to like.

It’s like people prefer that to a 2000-word blog post at this point. – Do you want to take a guess at how much traffic?

So this is SEMrush prediction, which is higher than AHRIS prediction of how much traffic has, but how much traffic per month do you think SEMrush estimates that Reddit has?

This is organic traffic only. – 750 million maybe, something like this. – 1.2 billion. – Yeah, okay.

Almost double. – That is, I can’t remember the last time I’ve seen an organic number that high in any SEO tool. – It’s like YouTube or something. – Yeah, it’s crazy.

So yeah, it’s like, can they lose traffic?

Yes.

Is there something to replace them?

And I think people like the Reddit widgets.

I use them now.

It’s like I got used to it.

Yeah, I think they’re staying where they are mostly, but the search playouts might change a bit.

Or AI overviews become even bigger, and that leads to less traffic to Reddit, but not because they’ve been demoted in a significant way. – Do you agree that Reddit has been inundated with spam and these people bumping up old threads with commercial content and buying uploads to rank?

And is that not kind of shitted up the platform a bit? – It does, but so does every platform.

We’ll talk about some other platforms later.

I’ll show you the state of Pinterest, for example, and other platforms.

And it’s a fucking shit show everywhere.

Instagram is full of fake AI profiles.

Pinterest is full of that too.

Medium is full of that. – YouTube’s even.

There’s a lot of AI, faceless channels. – So yes, it’s bad, and you see it more because they rank on Google.

But it doesn’t mean that elsewhere is better.

I mean, websites have been spamming this AI content.

I’m sorry, but the people who say that, they’re the same people who throw a bunch of AI content on their site themselves.

So you could argue it’s the same problem.

And it’s not like athletes haven’t been biased forever as well.

Let’s be honest. – I think what’s interesting and what maybe makes it less likely that Google removes Reddit so much is that Reddit takes one, sometimes two places in the SERPs, which means it gives Google plenty of other space to put all its other features, its AI overviews, has many ads at once and still nine other answers. – And it changed Gemini as well. – Whereas if we compare it to when there was a lot of small, medium-sized affiliate sites ranking or content sites ranking, that was taking up literally all of the organic space or most of the organic space.

So it felt like a bigger change when those were removed. – Yeah, it’s like now, you need to think about the SERP layout rather than like websites.

Like what is a SERP?

The SERP is now informational content is on top inside the AI overview and you’ll get the link to the sources to the content sites there.

That’s where they’re going to leave maybe one or two spots in the organics and that’s about it.

Then you’re going to have your UGC block, which is like the forum and discussion thing that mostly is Reddit and Quora at this point, but a lot of niche forums also do well there.

And then you basically have the commercial stuff all around and Google has this incentive to put Reddit for an informational piece over a website because it trains Gemini.

And we’ll talk about that because one of my prediction is related to Gemini.

It’s like, yeah, it’s something that’s very important to them.

This is training AI overviews.

This is kind of like a closed loop where they send traffic but they get something from it.

Whereas when they send traffic to your website, they don’t get much for it unless you run ads on your site that is run by Google. – Okay, so Reddit not going to lose any traffic or not the significant amount of traffic? – No, they might lose some, but I don’t see like a huge downfall. – Okay, you think what, 10%, 20% max? – Yeah, they could lose 15, 20% but no more.

That’s the way I see it. – Okay.

All right, so let’s move on to our next one, which is going to be around search GPT and perplexity as a viable Google alternative.

But first, I’d like to have a quick word from our sponsor today.

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.

Okay, so now let’s get back to the episode and we’re talking about search GPT and perplexity as viable Google alternatives.

This prediction comes from Nathan Gotch.

Agree, disagree.

I agree.

I mean, you saw in one of the previous episodes my default search engine was perplexity and my other browser is like search GPT.

And to go to Google, I type G and then I just type the query and it just gives me a Google query.

I think in general, search is gonna fragment.

It’s already more fragmented than people believe.

First of all, like 33% of all Google searches are now brand searches, so it’s just navigational.

I wouldn’t even call it like SEO at this point or just for the brand itself, but you’re not, it’s not fair game to try to rank for that.

I think, yeah, I think also Gemini is also going to get big if they put it in Chrome.

Like it’s quite possible that like the Chrome address bar gets some kind of Gemini search feature as well or like some kind of sidebar.

Assuming they keep hold of Chrome, that is.

Assuming they keep Chrome, I think they could take, they could eat into their own market just to prevent search GPT and perplexity to take it over.

As of the end of last year, search GPT is actually free to everyone.

You just need to log in and you can also make it your default search engine on chat GPT.

So I imagine like some people will switch to that.

So yeah, there will be viable Google search alternatives for searches, not for websites.

That’s another story.

They will send a lot less traffic than Google has.

And I imagine that a lot of people are going to miss the Google 90% market share days when this market share finally shrinks.

It’s interesting.

I’m looking at the way I’m using it and the fragmentation, that’s a really good way to describe it because I also switched to having perplexity as my primary search engine by default.

And I did the G to go to Google, but I found what I was doing is I was using Google more often, but for like a few of the same things, looking at the weather, looking at map for map information or like local search stuff, or for looking at like stock prices or travel stuff.

So I find like I’m kind of, I’m using Google, but I’m not really using it to find organic websites.

Yeah, I’m using it to find information that is displaying.

So it’s kind of becoming a portal in that sense.

Yeah, 60% of searches are no click searches at this point, which means like people do not click on any new result, which it means they probably rely on the AI overview or the widget, like the weather widget, whatever that kind of stuff.

And then yeah, another 34% of searches, they do overlap though.

So they’re a bit careful, are branded.

So it’s like, yeah, Google to go on websites is becoming a minority use of Google as a tool in general.

But this is really interesting because I think it gives small or very small companies like Perplexity, I mean, relative to Google at least, a real opportunity because they can essentially become the place to go when you’re looking for information about current events and to understand news and trends and these kinds of things.

So it’s really, really good at rounding that up information, presenting it to you much.

I think Perplexity is gonna have a rough time now that search is free.

We’ll talk about Gemini later, as I said, it’s getting good, like seriously good at search as well.

I think these guys, they’re just gonna bully everyone out of the market with their size and make everything free to basically buy market share.

Pretty much.

And so like Perplexity is like okay when it’s free, but like there is better alternatives now.

Like the paid version of Perplexity is pretty good, but it’s 20 bucks per month.

So it’s like, you know, you’re better off paying for chat GPT where you get all of the tools, including, I mean, including search in one package.

So it’s like the value proposition of the company is declining over time.

And it’s like, I think their best bet is to just get bought out really soon.

And, you know, they have made good money basically.

I don’t see them staying.

I think like all big companies have understood that search is going to be one of the main uses of LLMs and it’s going to be productized a lot more this year.

I think though the monetization model within that is still a bit unclear, right?

‘Cause I mean, even Google has this problem where they throw a bunch of AI overviews at the top and like, can you put ads in there?

How do the ads influence it?

Like that’s still all a bit of an unknown at the moment. – It’s not even important at this point.

Like the battle for market dominance is what matters and then monetization will matter more later. – Yeah.

All right.

So let’s move on to the next one, which again comes from Wes McDowell and this one’s around LinkedIn and specifically LinkedIn newsletters.

So, I mean, let’s talk about LinkedIn first of all, because I have found last year, it felt like something fundamentally changed and how people were using it and how popular it was, at least in a kind of B2B sense.

I know people were using it before, but like something about it changed.

Do you got a feeling? – Yeah.

I mean, it got more popular over time.

I think the downfall of Twitter has a lot to do with that.

A lot of people have migrated from Twitter to LinkedIn.

Like Twitter is now becoming a niche social network.

Let’s just say that to not be too involved politically and LinkedIn, like it used to be a network where a lot of professionals would share what they do, much more than like on Facebook or even Instagram.

And since Twitter is becoming like a more geared towards one specific target market, people that are not the target market and people who just don’t want to take political sites are moving their content to LinkedIn.

And I think that has helped significantly in terms of getting people involved. – I think there has been a lot of talk about, you know, left and right and blue sky and X and all this kind of stuff.

But I think what not so many people were talking about is the B2B crowd, right?

They don’t really care so much about any of that.

They just want to talk about work and it’s- – They don’t want to change it as well. – Twitter X is not really a place for work stuff anymore.

You go on and it’s like quite, you know, differing, different kind of personal videos and politics and that stuff.

And it’s just, yeah, not what a lot of people signed up for.

So I think that’s maybe helped.

But also on LinkedIn, they’ve been driving some pretty wild discovery, like ability for people.

So, you know, you create a good post on LinkedIn, even if you have minimal followers, you know, it can get a lot of reach.

And I think that’s how a lot of these networks in the early days build up users is by giving out, being very generous with reach, right?

We saw that with TikTok when it first came out and then they kind of dialed it back a little bit later on to try and- – And the reach is valuable, right?

It’s like business professionals.

These people who could become your clients or spend money with you.

Like it’s not like kids scrolling on their phone or whatever.

Like it’s people who have buying power.

And so even if you get a bit less reach, like the potential value of that reach is much higher.

And so like you combine the value of the audience with essentially the potential free reach by making good content.

It’s a good platform to be on.

It’s definitely one I want to spend more time on.

Like I used to be quite big on Twitter.

I’m not really into the vibe that it’s become to be honest.

So I think I’m gonna spend more time on social media this year and I’m gonna be spending quite a bit of time on LinkedIn actually. – Yeah.

So one I’m most excited for certainly at the moment as well.

And specifically, and this goes back to Wes’s point is they have this newsletter feature, right?

Which is kind of like a, I guess if you want to compare it to Facebook, it’s like a page, right?

You kind of have a blog on there, but it’s called a newsletter because on LinkedIn, they will send it to people’s email or LinkedIn inbox.

So you’ll get like a notification, potential email about it as well.

So it’s kind of like delivered in that way as well as being delivered in the feed.

And apparently like this is all the rage right now and is set to really take off this year.

So do you think that’s something worth investing in?

Bearing in mind that you still don’t get everyone’s email addresses to do other stuff.

It keeps you very much on the platform. – I mean, think about the game of these platforms, right?

The game of these platforms right now is like make posts that get lots of rich and then essentially sprinkle CTAs for people to go sign up for your newsletter outside of the network.

Like that’s pretty much all of these platforms now.

And really newsletters have replaced kind of like old school blogging.

Like, you know, when you talk about the stuff you’re doing and so on, like you’re more likely to do it over a newsletter these days than you are to do through a blog.

And so I guess LinkedIn has identified this behavior and was like, well, what if we bring all the newsletters internally and so like people stay on the platform, which is what’s important to them so that they can sell advertising for a lot of money.

LinkedIn is very expensive.

And yeah, I think that’s what they’re trying to do.

So these networks, they tend to reduce your reach when you have links that point externally, especially to sign up to newsletters, et cetera.

And now they’re like, well, maybe you can maintain your reach if you actually run your newsletters through our platform.

And so we keep full control of that.

Now, the problem is these things, they tend to work for a few years and then they kind of like get less popular.

It’s more difficult, like LinkedIn might change.

And so like if I was serious about building a long-term business, I’d still want people’s email addresses.

And if I’m unable to collect them, then I’d probably be willing to forego the extra boost in growth.

Or maybe I would do two layers, right?

I would do a LinkedIn newsletter.

And inside the newsletter, I would have CTA’s to sign up for stuff inside, like maybe extra lead magnets or something like that on my website so that I get the extra reach from the LinkedIn newsletter.

But I actually eventually do send engaged people into my real email list so that when LinkedIn newsletter inevitably becomes less popular, I still maintain my reach through my email.

So like, is it gonna…

I see it as a temporary gimmick and a way to grow your real email list maybe.

Is it like a long-term thing?

It’s like, there’s a lot of these features on social networks and they tend to not last super long.

You’re right though, it’s so important to get your audience off platform and have some kind of connection with them in another way, through your website, through your email list and in another way.

Because these social media platforms, that is their game.

They’re your friend for ages, they help you get loads of reach.

And then they change something, engagement dips, reach dips, you have to pay for things and yeah, you lose out.

So there’s not so much long-term stability with…

But there’s ways to leverage these temporary features that work very well.

This newsletter, if LinkedIn is really pushing them, like a lot of social network pushes new features like a lot, then I would do that.

I would make a LinkedIn newsletter that just essentially promises free information.

And then I would just put call to actions to lead magnets inside these newsletters where people have to click through and I know tin.

And that’s maybe an extra step, but if I get free reach in exchange, it could be worth it.

So yeah, it’s almost like it’s not really a newsletter.

It’s kind of like a pre-step to the real newsletter.

If it gets big, yeah, maybe.

To be honest, I don’t know that features.

I’m not going to pretend I know what’s going to happen.

But LinkedIn is definitely a place to look at.

So yeah, maybe it gets big.

All right.

So let’s move on to the next one.

This comes from Jade Beeson.

She’s a popular social media influencer.

She talks a lot about things like the future of social media and using AI in social media.

Her…

Okay. …point is that we’re on this kind of like we’ve reached this tipping point in terms of creativity.

So there were huge barriers before to making like video content and especially like well, well produced.

Your ability to take a creative idea you had in your head and make that a reality on the screen, it was difficult unless you were kind of a professional creator.

So her argument is that with tools like Sora and you know, video AI, video generation things and just AI in general to help brainstorm and develop content ideas, the average business owner is going to be much more able to bring their ideas to life quickly while still focusing on their business and that kind of thing at the same time.

So do you think we’re a tipping point here or not?

So are the tools going to enable us to do more and better things?

Yes.

Are people going to do more and better things?

I’m not sure.

The problem is…

Are we going the other way then?

There’s just more things but not necessarily better.

Yeah.

It’s like, I mean, look at how it’s gone with AI text generation, right?

It’s like, everyone was like, “Oh my God, we have these great like we used to pay writers, you used to take weeks to make articles.

Now we have these tools that can do the same thing.

Did we create better content?”

I don’t think so.

So I think what happened is…

For most people, I think.

Like 5% of people did, but 95% of people…

They got drowned.

Spammed it to shit.

And then all the platforms were like, “This stuff is terrible.”

And they stopped showing it.

So do you think the same thing, what we’ll see a ton of videos or short form video content is just crap and then the platforms will get much tougher?

Or we’ll use, you know, user engagement metrics.

Exactly.

Social media networks have, be a better filter of that than the Google algorithm was, for example.

Yeah.

So the end result is the bar’s gonna rise because essentially the people who do the better stuff still, as is now, get most of the reach and then the bottom will get better quality, but still be the bottom and still get the same reach.

So I don’t see anything change fundamentally.

The one thing that changes is that if you want to do quality and maybe like, for example, I see it for video editing.

I’m kind of getting into video editing with like CapCut, et cetera, right now.

And it’s like, you can do crazy stuff that I could never do before with this.

Like, for example, like, you know how YouTube, YouTubers, they have this background where they have like, you know, a red light and a blue light, et cetera.

And it takes quite a bit of setup to do, right?

I could do this on the video I’m doing right now, even though there’s just a white background.

I can relight it.

It doesn’t just put like some red and some blue.

It relights my face and every item using AI, pretending there’s a light source that’s say on my left and a lot of sources that say on my right and it calculates the crossover of the light, et cetera.

And it’s like, it’s not perfect, but it’s good enough for like short form content and so on.

And so like, it’s going to allow you to do these things.

But if you’re someone that has no taste or nothing to say, I don’t think that’s going to help you very much, what it’s going to do is just going to rise the bottom.

Like, it’s going to be expected to have that, you know?

This is the trap with AI.

And we saw this with AI text, and I’m sure we’ll see it with video as well, is that because it can string a sentence together in a really nice sounding way, or because it can make a video look really polished, it doesn’t mean that the underlying information or content of the video is good, right?

And I think that’s where people have to focus on if they want to be successful, regardless of what it looks like.

AI has no taste.

And if you don’t have taste to say what’s good or bad, you will make crap content by overly relying on AI to create that content.

And that content is going to fail regardless of how good the tools are.

So yeah, the bottom is going to rise.

That’s the way it’s hit.

Okay, so the next one I have here comes out of exploding topics.

And they’re predicting that influencer marketing platforms specifically are going to be a big thing in 2025.

So influencer marketing, the classic example, a business wants to promote a product, so they pay an influencer who has an audience to talk about it, to share it.

And the way they connect with the influencer, it’s always been a bit difficult.

The pricing’s been opaque and you’re not really sure who you can contact.

But there are starting to be established tools in this space or databases and agencies and things like that.

The whole ecosystem is being built up.

So they mentioned specifically Upfluence, which apparently has over 3 million influencers at the moment.

I haven’t used it myself.

But what’s your thoughts?

Influencer marketing in 2025, is it going to go big?

Yeah, I think so.

I think that’s where kind of like a part of a fair marketing is going.

The part for small creators and so on.

Brands, they’re much more willing to now invest money into creators that have audiences and essentially promote the product without cannibalizing on their brand searches or whatever.

They’re much more refined in terms of how they can create their conversions.

And they see that all the way they run their affiliate program doesn’t necessarily work.

And it was showing conversions, but it was not necessarily people who would have not converted if they had not come across the affiliate.

And so as a result, brands, we’ve seen it with the podcast also.

And here, it’s something that has grown for us as well.

Yeah, it’s going to grow.

People will want to invest in audiences, especially as social networks are going to become more saturated with AI content and people’s attention is going to be scarce, basically.

I think there’s this perception or there has been that an influencer is one of these big YouTubers, someone with 10 million subscribers and all that.

But actually what’s happening is that there’s a lot more of these.

So there’s a term like micro influencer and now a new one, nano influencer, which is like someone that has less than 10,000 followers.

But we noticed this ourselves when we were first growing authority hacker.

The first few thousand people in your audience, right?

You have almost personal relationships with a lot of them.

You email them, you interact, they comment on things, and you kind of get to know a lot of these people.

Once you get past of 10,000 and it’s the hundreds of thousands, it’s really, really difficult to keep track of ever.

There’s just so many people in your orbit, right?

So you don’t have quite as strong of an individual relationship.

So I think the argument is that these nano influencers can in fact be more influential towards a small specific subgroup of people.

And because their audience size is less, it’s cheaper for brands to work with them.

So it’s almost like a better deal in a way to work with those people.

And in the B2B space, it’s kind of exploding right now.

Yeah.

And I think Neil Patel showed some stats around that where he essentially showed that there’s a much higher ROI in terms of investing a small amount of money with a micro influencer than just paying very, very big celebrities with millions of followers.

It’s also much more achievable for even small brands to use that as a marketing channel, because it’s like you’re talking a few thousand dollars at most for most engagements.

And so, yeah, I think it’s going to grow.

We’ve seen results.

We’ve seen people having results with that as well.

And I think where the kind of genuine marketers are going is going to be that actually.

And the money is going to shift for the people who build audiences.

So some statistics, according to exploding topics, influencer marketing is growing three and a half times faster than social media ad spend.

So companies already spent $2 billion on influencers in 2023, so two years ago, and expected 65% of businesses are planning to increase influencer marketing spend in 2025.

That’s a lot.

Growing 3.5 times faster than social ad spend is a really, really strong trend.

Which itself is growing, yeah.

Yeah, exactly.

It’s not a declining market.

It’s like I think there will be a lot of money there.

And it’s like a lot of people who have been struggling with Google, etc., would be well-advised to essentially investing growing audiences across all platforms and consider this as a monetization channels.

And the caveat is you actually have to do the thing you’re talking about, because when you start doing videos and so on, there’s no way you can fake it.

Yeah.

All right.

So let’s move on to the next one.

This is the last one of our predictions.

And this was kind of mentioned by a number of people.

From the guest predictions.

Yes.

And then we’ll get on to our own ones.

But this was mentioned by a number of people.

So it’s around AI chatbots being used for customer service for online companies.

And apparently 72% of customers expect an immediate, which is defined as within 10 minutes, response to their customer support queries 24 hours a day.

So surely AI chatbots are the only way you can meet that demand.

But are they ready?

Yeah, I think for most, the way I see it is like you get a low level of support handled by chatbots.

And you can teach the chatbot to escalate to a real human when it’s something that becomes a bit more complicated.

I actually believe that’s going to happen with doctors.

So that’s the way I see it.

I can see the NHS and all the social healthcare services that are overloaded with patients essentially handle all the calls and small stuff with AI chatbots and just pass on the important stuff to doctors.

So customer support, definitely.

I think that’s going to be great.

I think doctors will be one of the last ones that do this.

I mean, knowing what I know about the healthcare system in this country, it’s quite archaic.

And to suggest that AI will take over, I mean, I find it a bit of a stretch.

You know what’s going to convince them?

Money.

Like it’s going to actually kill the deficits.

It’s a fair point, actually.

I think that there’s huge wait lists in most Western countries to see a doctor, even if you’re paying money for one.

This is a way to fix that.

And we can get into this.

Let’s not go into that because we’re talking about customer support.

But my point of this is that I think that this will be the last area that AI takes over because even though you can maybe prove that it’s like with driverless cars, like you can prove they’re safer, but when they do make one mistake, even if that mistake percentage-wise is much less than what humans would make, people can’t handle that a machine kills.

I understand that the NHS has never made any mistakes.

So it must be better with AI.

Anyway, let’s go back to customer support.

Yeah, I think that you have two layers.

You have the lower layer.

You expect the lower layer to be AI.

It’s already happening.

I wish ActiveCampaign did that, by the way, because it’s been three days I’m waiting for a ticket response right now.

So ActiveCampaign, please check that out.

But then you have a lower layer.

And then after that, it would just escape.

It would be given a list of cases where it should escalate.

And be like, “Hey, I’m passing this off to a human.

We’ll be back to you in 24 hours.”

And then you’ll have to wait the normal ticket time.

Which means probably the support jobs are going to decrease a lot in 2025.

There’s going to be a lot less people behind a lot more tickets.

And the thing as well is with a…

Sorry.

This is a classic example of it’s a job where 50 to 70% of the tasks are basic and repeatable.

So like that segment.

And also the consequence of getting it wrong in most industries, barring healthcare and a few others is minor.

It could usually be fixed later.

So that is absolute prime for an AI takeover, I think.

Yeah.

I think that I was going to say as well.

Now, we have these models where you can show your camera and you can show all that stuff.

Gemini 2.0 Flash is natively multimodal with video.

It works decently.

I can imagine that you have a problem with your car.

You have not just software, like sync everything.

And you’re just going to point the camera and talk to the person and they’ll just help you with that or share your screen or whatever.

And that’s going to almost be better than the traditional ticket system that you have with support agents.

So there’s a case where it’s going to upgrade support for many products while also making it significantly cheaper.

All right.

So let’s move on to some of our own predictions for this year.

Gael, do you want to take the first one?

Let’s maybe do two, two or three each.

All right.

So the first one I’m going to take, which is a very controversial one, is that Google is going to take the lead in the AI race, actually.

Okay.

Quite controversial.

Yeah.

But I wouldn’t have said this six months ago, right?

No way.

But they’ve just done releases at the end of last year, the new Gemini model.

So there is the Flash model, which is kind of like the less smart one, the smallest model.

But it’s already as good as their mid-sized model.

So the upgrade is significant.

And it has the biggest context size by a mile.

Just to give you an idea, these models from Google have 2 million token context size.

Consider tokens like a word, basically, something like that.

So you can put multiple books inside one prompt and it can handle all of that.

I think the demo from the 1 million context size was they put Le Miserable from Victor Hugo.

It’s like a French book, pretty big book.

And then they uploaded a painting of a scene inside the book.

And they asked the model to go and find the passage in the book that matches the picture.

And the model was able to do that.

Now we’re talking twice more context size.

So you could upload this book more than twice.

Can you give us an idea of what current context sizes are in some of the other platforms?

Yeah.

So GPT-4 is 128,000 tokens versus 2 million for Gemini.

Entropic cloud is 200,000.

So we’re still talking like 10 times more than Entropic inside the Google models.

Yeah, awesome.

They’re leading on that.

And it becomes important for things like imagine like we talked about customer service, right?

You can give a lot of data inside the prompt from your customer.

You could put your entire customer service manual in there.

You don’t even have to build– Yeah, you can upload hundreds of fast queries and analysis of what and what.

Exactly.

Many, many replies.

You can make a huge– it makes this better.

They have a free API.

You actually know that the Gemini API is actually free up to a certain rate limit.

You don’t pay for it.

And when you pay for it, it’s probably cheaper than everyone else.

So I can imagine a lot of apps– Basically, the models are slowing down in progress right now.

Everyone’s like, OK, we ran out of data.

We need to figure out what’s the next breakthrough so that we can make a big leap again.

And so what’s going to probably happen is the next generation of models, they’ll be more or less the same level of smartness.

So price is going to matter, right?

It’s like how cheap is it and how fast is it?

And Google is leading on the price as well.

So they’re not only leading in the context size, they’re leading in price.

But one thing that makes Google’s models particularly interesting is that they can integrate them in the products we all use.

So for example, the API has an option where you can route it in Google search.

So think like Perplexity, but via API.

Like Perplexity has an API, but it’s a piece of shit.

So but this one is actually decent.

So it can base the answer based on what it finds in Google.

And it’s just a tick box in the API.

They’re building Chrome extensions for it.

What would the use case of that be then?

I mean, I’m not sure, but let’s say you want fresh news, for example.

Like most of these models have like a cutout date.

Like chat GPT has search, but it doesn’t have search inside the API at the point at which we’re recording at least.

Whereas here, like you could route it into like, okay, what happened yesterday?

It can do a Google search and then have an API answer.

Oh, then work with that data.

Okay, exactly.

But like, you know, Google Maps, for example, like it can make itineraries inside Gemini.

So if you ask it Gemini, when it’s an itinerary, they will show you like a Google Maps preview and prepare the itinerary.

And you can click it, open Google Maps, for example.

They have access to all the YouTube data.

They have access to Gmail.

They have access to all of that.

And so like once they already started baking a lot of that stuff in there.

And I think, yeah, it’s going to just make it a better product, even as the eyes are just not really much smarter than each other at this point.

And also, I think Google search is going to learn from search GPT that is gaining traction right now.

And just basically just implement whatever is winning there, like whatever people like.

Like, for example, they at the end of last year, they already started testing with you being able to drag and drop a file into the search field for Google to search in it.

The same way you would do it in GPT, for example.

And I can see these things essentially happening more and more as OpenAI challenges Google search.

They identify what works.

But yes, in terms of AI.

Is this going to be like the big social media companies where they all like someone innovates a new feature and then everyone else copies it the next month?

Of course.

And so, but like Google has the usage.

And so if they’re able, and now that their models are really catching up, and they’re cheaper, and they have more context size, unless there’s like another leap from the other companies, which can happen.

Like, you know, I can still see Entropic having the smartest models, for example, like Claude is it’s still probably going to be there.

But it’s really for professional niche, et cetera.

It’s not for everyone.

I think the model everyone will be using could be a Google model by the end of 2025, if everything goes well.

Okay, that’s an exciting look.

Watch this space, I guess.

Now, that’s all assuming that Google doesn’t get broken up by the government.

And that could happen.

They’re all spanner in the works.

And that’s kind of my prediction.

So there’s two stories that are kind of ongoing at the moment.

One is TikTok and it’s imminent ban in the US.

And the other is the Department of Justice wants to break up Google, essentially, and make Chrome a separate entity or sell it off in some way.

It’s not quite clear yet.

But both those things are supposed to happen this year in 2025.

I don’t think either of them will.

Both of these are kind of too big to fail in a way.

They make so much money and they’re so powerful that they will fight tooth and nail to have some kind of compromised solution.

Kind of like what we maybe saw with Microsoft and its antitrust case 20 odd years ago, where they just dragged it out and dragged it out to the point where it was unimplementable, basically.

I don’t say it’s not just that.

It’s like if you sell Chrome to anyone, it’s Chrome’s like 15 billion valuation or something.

Right.

So not many companies can just afford to bankroll that.

If you give that to Facebook or Amazon or Apple, you create instantly another monopoly that is just the same issue that Google has.

You know, and you know, I think these governments, especially in search, they’re like talking about like they’re four years too late.

Right.

That would have been the right thing to do in 2020, 2021.

But now we’re in this new era of AI and like Chrome’s now factor in the AI arms race and they’re not really considering those things as well.

So they’re they will if they do break it up, they’ll tinker.

And I don’t think they will make the problem any better.

It’ll just go to someone else.

Right.

Yeah.

It’s just like I think you create an instant other monopoly for pretty much anyone else that buys this.

Can you imagine Amazon buying these?

And you just see like a bunch of you have Chrome Prime or something like, I don’t know.

How does that work?

Yeah, that wouldn’t work.

And you know, the TikTok one is slightly different.

It’s on the kind of national security grounds.

And you know, I get where they’re coming from.

It’s just so popular now, TikTok, though.

I think it’s going to be difficult to actually.

Trump wants to protect it as well.

So we’ll see.

We’ll see what happens on January 20th.

But yeah, it’s like I think the ban is supposed to be enforced on January 19th, which is really soon.

Yeah.

And Trump comes into office on January 20th.

So that should be an interesting one.

Even if it is banned, it’s not like, you know, everything gets deleted on the first day.

It just like it will be unavailable from the app store.

Current users can still update it.

So like the new government coming in, if they want to 180, like they can do that.

No problem.

Yeah.

Trump said that it helped his campaign, so he liked it.

That’s what he said on the interview.

There you go.

So we’ll see what happens.

You want to take the next one?

Yeah, I think that next year we will be spending a lot of time tinkering with building autonomous agents.

Now, this is something that’s already starting.

And actually, like I’m going to share my screen and show you that.

So if you’re on the audio, I’m just going to show you guys.

So actually, we already have AI that can use your computer.

So I’m going to play this without the sound because otherwise that’s a bit much.

But basically, you know, that’s cloud that you’re seeing here.

And yeah, describe what we’re looking at for the audio.

Yeah.

So this is a cloud computer use demo where they essentially ask you to fill out a vendor request form with some equipment.

If it has in a so sorry, with some data, it has in a spreadsheet.

And you know, the model basically reads the.

So we’ve got a Google sheet opened in a Chrome browser, and then we got another Chrome browser with like a form from a website.

And it’s like essentially going through and acting like a human performing.

Yeah, that’s getting the information from one place, copying over, pasting in the form, going back, copy pasting it like it’s I feel like if a user requires a macro of themselves, this is kind of it performing, but it’s all led by the AI.

Yeah, exactly.

And it’s like, remember, models are becoming multimodal now.

So you’ll be able to upload a video of you doing something.

And that thing is just going to do it again and again.

So like, you know, I don’t think it’s going to do like crazy, you know, crazy jobs or anything.

It’s just going to do things like lead gen, sending basic key data and I can still.

Yeah.

Oh my God, you must be worried if you’re in that industry.

Yeah.

And Google is doing the same thing with something called Project Marina, which is a Chrome extension based on that new multimodal model.

And again, you have a chatbot on the right.

And so it’s like, I’ll memorize the list of companies and look them up and find their contact information, that kind of stuff and fill the spreadsheet.

And essentially it does it now.

It’s very slow again, but it does it like I’m skipping through now.

But you can see it’s like it’s going through websites.

It’s finding it’s going on Google is searching on Google, going like finding the contact info, going to the spreadsheet, adding it, etc.

It’s essentially doing it slower than a human at this point.

But we would have to assume that that would dramatically change, like quite quickly.

Also, like you can launch this at night and just wake up in the morning to a filled up spreadsheet.

Like you don’t have if you don’t have to like monitor it and just let it do its thing, you launch it and then you come back a few hours later and it’s done a bunch of work.

So I think that’s going to replace like, you know, the chief VA in the Philippines or something like this, like this kind of job.

I don’t think it’s going to do much else in 2025, but it’s still something.

And we’ll spend quite a bit of time, I think, building this.

How big a risk is there that this gets into the hands of some scammers and it ends up sending money to different people’s bank accounts from their logging in and doing all that stuff?

99.9%, but I’m sure the reason why it’s not…

I’m sure the ScamBait or YouTube channel is going to have a lot of content in the next few years, I guess.

Yeah, I mean, it can be used for good and for bad.

It can be used to spam people.

It can be used to like, really, it can do anything a human can do.

Humans do good and bad things.

Yeah, I think it’s not going to be much smarter than that, but I think there will be a lot of like very useful use cases like, you know, lead generation.

People are like doing some basic research, finding like, hey, find me companies that do this or whatever, and that’s going to be pretty good.

Talking about finding companies that do this, another thing I didn’t say about Google and about their model and winning the AI race is they have a new thing called deep research.

And it reads literally like hundreds of websites and gives you like a giant Google doc research on whatever query you gave it.

So, it’s in perplexity on mega steroids.

You give it a query and it goes like an intern that researches for a week.

And an hour later, it gives you a Google doc that has like super in-depth research on everything that you wanted to do and brainstorm multiple other queries, related queries, etc., read hundreds of documents and then made you a summary in a Google doc, which is crazy.

What a time to be a student at the moment.

Like you have so many tools to like do your work for you.

To do nothing.

Can you imagine how much they can party now and like not care about being hanged over the next day?

That must be fun.

Yeah, those are the days, hey?

So, that’s that.

And then there’s something that’s a little bit less new, but that’s the real-time API.

Real-time API is like advanced voice mode on charge GPT basically.

You know, the one that you can interrupt, that you can talk to that has low latency.

And I think if you combine, you know, these agents with essentially real-time API that, you know, can do phone calls, can do all of that, etc., we’re going to build some pretty crazy tools.

Like I can imagine having an AI chat that I talk to like, hey, book me a hairstyle list every six months, put it in my calendar and do this and that.

And you’ll be able to do that.

Now, the only problem with this is still…

Hold on.

Hold on.

Yeah.

You get your haircut every six months?

No, I mean, okay, no, I don’t get it every six months.

I just, I was going to say something else, but I just said, yeah, I got it.

Okay.

Let’s talk about the price though.

The problem with these APIs is they’re expensive.

$0.24 per minute of audio output, sorry, on openAI, which is more than an Indian call center worker at this point.

So it is still pretty expensive.

I expect these prices will probably reduce in 2025.

But yeah, I think we’re going to be building some agents combining all these things next year.

And it’s going to become even more of a game of optimizing your processes, templates, and just getting things like, getting these things work for you.

And I’m pretty excited for that.

That’s one of the parts that I’m the most excited for next year.

Okay.

Let’s take one more of our own predictions, and then we can go up and review last year’s one.

So you want to take the last one?

So that’s what I’m going to take is actually on social networks, just becoming just as flooded with AI content as we’ve seen blogs become, right?

It’s like we have a lot more AI models that can generate really good images.

We now have video models like the Sora that came out that wasn’t nearly as good as people expected.

But actually Google has also one that just came out.

I can’t remember the name, but it looks better than Sora.

But Google seems to be losing the kind of PR war against openAI.

Yeah, they’re telling you it’s good.

Hardly anyone has heard of any of these things that Google has released.

Maybe.

Maybe.

That’s good in some cases.

Yeah, they should call us maybe.

But yeah, it’s like it’s like it’s a good one.

And yeah, it’s like there’s already a lot of social networks that are like flooded with AI that nobody talks about.

But I think with this stuff, it’s going to be like it’s going to be like blogs really.

And I want to show you again what Pinterest looks like today, because I think that’s quite interesting.

So let me share this with you.

This is my homepage on Pinterest, like nothing crazy, except this is an AI image.

This is an AI image.

This is an AI image.

This is an AI image.

This I’m not sure it may be an AI image.

This is an AI image.

We’re looking here for the audio listeners, just pictures of people’s houses and gardens and bathrooms and offices.

Usual Pinterest stuff.

But yeah, has that kind of like 70% of it AI is AI.

Yeah.

Like and this is just like, yeah, so this is probably real, for example.

But this is AI.

This is AI.

This desk for sure.

Like when you zoom in, like this is I don’t think that’s real.

And you could keep going.

Like it’s like, I would say like 50 to 70% of the images I have on Pinterest are AI at this point, which is bananas.

And the same is happening on Instagram as well.

For example, where people are making first step with fake hot girls and just putting an only fan.

And then it’s like people teach that.

Like it’s not, I was researching for this podcast.

I saw tweets of people like selling courses on how to do this, you know, making fake AI models, attaching an only fan and a brag about making 20K a month, basically.

Right.

And it’s like, I was also playing with custom training AI models on my photos and essentially have it come out, like see how close I could get to faking it.

Right.

And I’m going to show you two photos I generated.

I showed you the funny ones earlier, but I made some more realistic ones.

Look at this one, for example.

It’s like he’s a little bit more in shape than I would be probably.

Like, I wish I could look like that.

But other than that, Maybe a bit better fashion sense as well.

But yeah, maybe other than that, AI dress is better than me.

That’s a fact at this point.

I mean, so I can’t remember where that photo was from, but I feel like I’ve seen that like head pose of yours before.

Is it just taking that and put it onto like another body or a similar looking body?

I mean, it’s like, yeah, it’s like, I mean, it’s slightly more in shape than my body, but it’s not too far off, I would say.

So no, it’s just like, it’s just like, it’s just a prompt that was, I told it to like make it look like an iPhone photo, like not perfect, etc.

And just like, it feels like that.

And here’s another one.

This one is a little bit less good, I think, for the face.

It’s got that kind of like super like skin smooth AI thing to it.

You know, they need to be obscured.

Like there’s tools to like upscale them and make them better.

But the point is like, if I kind of like tweak this image a bit, I’m pretty sure I can post something there and people are going to think it’s real.

Like if they’re not really looking for it, and if I’m not really caught posting too much AI content, and the point is like, yeah, there’s tools to like upscale this and make this a bit more high detail and so on.

But like, yeah, this is a model that took me 20 minutes to train on my photos, and then just a prompt.

And then I can generate all sorts of crazy images.

Like, yeah, let me actually just see if I have some other ones.

But here’s me as a Jedi, for example.

You can see.

So yeah, you can do some crazy stuff.

And I think social media is going to be flooded by it.

And I think that’s going to lead to maybe even some platforms making human only content or something like that, because social media is going to suffer the fate of what happened to blogs and content.

And I think people are going to really crave for like real authenticity.

And that’s why some creators like Sam Soolake, etc. on YouTube, do so well because this AI content is going to be everywhere and tiring.

What’s interesting is that when you have the Pinterest page up for, you’re able to look through it and make a judgment about which photos you think were AI and which were not.

How far away do you think we are from a point when you cannot make that distinction?

Maybe two, three years.

Like, again, there’s like upscale tools that make things look a lot better.

And it’s like, I didn’t bother for my examples, but like, yeah, you can go one step above already.

And I think that’s what these fake Instagram influencers teach in the course, basically, if you want to learn that.

But yeah, it’s not…

The last 10% is always very hard to get.

It’s like, we’ve seen self-driving cars since 2016, right?

They still don’t self-drive in the car at this point, in the street.

And we would be quite afraid to let them self-drive for like a whole day, being in the car, for example.

So I think AI images will probably go that way, where it’s still going to be identifiable for a while.

But getting that last 10% is progressively…

But like, maybe now, like 10% of people can tell it’s an AI image.

And in like two, three years, we might be going down to 2% or 3%, you know?

Okay.

That’s probably what’s going to happen.

All right, so let’s move on to the last section, where we’re going to react to some of last year’s predictions that we made.

So let’s just go through them.

I’ll say what the prediction is, and then tell me, yeah, we got it right or wrong.

And let’s count on the total.

So the first is, we said that there would be a split in the SEO industry, and it would become more divided between two approaches.

One being those chasing quick wins and algorithms, and the other focusing on building quality brands through EAT, genuine UX signals, and just building a good business.

I mean…

You got it right?

I feel like we didn’t.

Like that’s kind of like… – I feel it’s scattered more than anything.

It’s more than two sides, I feel.

Like there’s just a million different churches at this point, and people just follow them.

So yeah, this industry is slowly dissolving in my opinion.

It’s just happening.

So it’s fair to say we got it half right, so we’ll give ourselves half a point for that, but it definitely has gone more than two ways.

There’s like a thousand different ways that it’s split up.

There’s still people chasing the algorithm.

There’s the people who pretend nothing’s happening, and everything’s fine.

There’s the people who left.

There’s the people who became a typical authority experts overnight.

There’s the people who want to sell your links.

There’s all of them.

And so yeah, at this point, I think many people are leaving this industry though.

It’s not something that’s growing anymore.

I think everyone can agree on that. – Yeah.

Yeah.

Time to abandon ship for everyone then. – Well, yeah.

It’s like that’s one of the things that we talked about when we said we want to kind of move on ourselves.

It’s like, “Oh, I don’t want to be the last guy at the party at 4am.”

You know what I mean?

Like the drunk guy on the dense floor, everyone’s left and holding a drink that’s spilling all over the floor.

Don’t want to be that guy.

So yeah, we may be biased on these though, so we will give half a point. – All right.

Next one was there’d be more ranking volatility followed by a period of stability. – What is the stability? – We thought that the first half of 2024 would be quite volatile, which it was to be fair.

You know, the core update, the Google core update in March through May was pretty wild.

Then they also announced the search site reputation abuse stuff, which kind of again dragged on into, I guess, Q3.

But it felt like the updates just kept coming and coming and coming. – I think it was stable about the bottom, you know?

People got the bottom. – It’s like it couldn’t get much worse, right? – It got stable.

So in a way we were right, but not in the way we expected to be, right? – I wouldn’t give ourselves any points for this though, because even at the bottom of the stability, it still felt like every core update they had, and there were a lot of them last year. – Even the December ones, they did. – There were sites getting wiped out.

Yeah, like in many cases, undeservably so again.

So it’s like they haven’t really fixed the core issue here. – I’m not sure they’ll fix it in 2025 either, to be honest.

At this point, it looks like more changes are coming due to increased competition.

So my guess is more of the same in 2025.

I have to make a prediction. – Okay, next one was Google will address big sites ranking for everything.

And I think we got this right, because they did nothing about this for decades.

And then in the second half of 2024, they did actually start taking action. – They went after Forbes.

To me, it’s like the biggest symbol they could ever have there. – Yeah. – So yeah, a lot of sites lost traffic.

I would not be surprised if many of them regained a lot of traffic in 2025.

But it’s not as free for all for the big sites on affiliate queries anymore.

And actually, in many, many cases, they took their rankings. – There were two angles to this.

The first was in May last year when a number of sites that had these coupon directories, which were apparently all managed by the same one or two companies just using the domain reputation of big newspapers and sites like Forbes to, yeah, rank for coupon queries.

So that got basically destroyed in May.

And then in, I think it was September, they went after September, October last year, they went after Forbes in this big profile thing like Forbes advisor.

The whole affiliate section of Forbes basically got taken down, got de-indexed.

And many other sites, CNN and big publications got affected too.

So that was a big wake up call.

And those publications are going to have to make major changes this year if they want to get any of that traffic back. – Yeah.

And a lot of people got fired as well.

It’s not just the small sites that took the heat here.

It’s like a lot of these big sites are under fire.

And it was kind of the ICU.

Let’s just say that.

Like it just came two years later, but it was the same. – Okay.

So let’s put that one down as a one point for that.

We got that right.

Next one, Parasite SEO will be addressed.

Google will tackle sites selling guest posts content for SEO purposes and may update their webmaster guidelines and implement manual penalties.

So I don’t really think we got this right because Parasite SEO is still very much alive and kicking. – But Auto-Kendia is not.

So you could argue that a lot of the sites that used to sell placements are not here anymore. – Okay. – They’ve got… – So I would say they addressed it like a kind of surface level and then it became fragmented, right?

This still happens a lot. – Yeah. – But it’s maybe a little bit kind of under the table and less in positive. – I think it’s more challenging.

I think the costs went up and the returns went down.

I don’t think it’s nearly as profitable as it was a year ago.

And I think that’s not too difficult. – I’m not sure though. – I see so many people talking about this on Twitter and SEO conferences. – Yeah, selling Parasite SEO courses. – Yeah.

I mean, it seems to be thriving. – Selling placements. – Yeah. – Be careful what you read. – I wouldn’t go as far as to say they fixed the problem. – No, they didn’t fix the problem, but they made it much more challenging.

I don’t think nearly as many people can profit nearly as much as they used to.

And that’s kind of like what they need to do.

They need to keep it in check.

I don’t think they can ever completely fix it. – Okay.

So do we get a point for this? – I would say yes.

You can put half point if you really want to be fair. – All right.

Let’s go half a point.

Next one, we said there’d be a major link related algorithm update. – That did not happen.

There was no. – There was not.

There was not at all. – No. – It looks like- – We did see more manual penalties, but it wasn’t really for link related stuff.

Mostly for content related. – In general, the traditional ranking factors, it doesn’t look like they’re touching them much anymore.

It’s more like they’re adding new ones and they’re user engagement, brand searches, stuff like that is coming in rather than Google is just touching on title tags or whatever.

Like it doesn’t, it feels like all this stuff is shrinking in importance.

It’s still there and important.

But it’s shrinking to make room for new factors rather than them changing how Google handles them. – Okay.

So let’s move on to the next one.

So that was SGE, search generative experience will- – Which is now- – Realize but won’t kill SEO. – So- – So, I mean, it rolled out to a billion people. – Pretty accurate. – Yeah, it rolled out to a billion people in 2024.

So that’s a lot of people.

It didn’t kill SEO.

No, it’s like SEO is almost impossible for content sites.

Like that’s something I want to say because when we made a lot of announcements and last year people got confused.

SEO is still viable for most real businesses like Econ, SaaS- – Local businesses. – Businesses, local businesses.

It’s still good.

And you can still get good traffic from it and make money.

It’s just the content site business that is not something that we would recommend people do anymore.

And that’s why we just don’t sell the courses anymore.

So it’s like, yeah, didn’t kill SEO.

It just killed one business model. – Yeah.

So I mean, it’s fair to say, I think we got that one.

Yeah, it felt kind of like a featured snippet.

You know, it’ll eat into your organic traffic, but not destroy it. – Yeah, and that’s exactly what it does.

It’s just like, it takes away clicks, but yeah, people still get traffic from Google. – Okay.

And the last one was, “GPT 5/Gemini content will make GPT 4 content feel outdated.”

Didn’t really get that one right. – We need to be honest at this point in the episode.

Only the real loyal fans are listening anyway.

It’s like, we’re recording this while OpenAI is making their 12-day announcement in December.

Therefore, there could be a chance that they release something like GPT 5 on the last day, which is speculated.

So I don’t know yet at this point.

We’re on day eight at this point.

But it’s actually not GPT 5 or Gemini that has made GPT 4 content feel outdated.

It’s actually Claude 3.5 Sonnet that has done so, which is a way better writer than all these other models by a mile.

A lot of people who attempted to create content with AI, that we helped, they used to use OpenAI, they struggled.

We made them switch to Sonnet and change their prompts and change all of that.

And now they’re like, “Well, I fired all my copywriters.”

Sorry for the copywriters, but yes, Claude 3.5 Sonnet is very, very good.

And I think that’s the model that really took creative writing to the next level this year.

So we were wrong on the model, but there is a model that made original GPT 4 content feel bad.

And that’s 3.5 Sonnet.

Okay.

Then I guess we get half a point for that maybe.

Yeah.

I think we get half right.

So let’s see.

We got one where we…

Half right where we…

Two, three.

We got three and a half out of seven.

So exactly 50%.

So basically you could flip a coin and just get the same accuracy, right?

Exactly.

So yeah, I’m not really sure if people should be listening to our crystal ball predictions in future.

But let’s be honest.

I mean, it’s difficult to estimate exactly what’s going to happen.

And this is fun to do.

We…

Something we do every year.

I think we get a lot of stuff that’s like kind of in the right direction, but not really turns out that way.

A lot of it is like the timing didn’t…

We’re predicting the general direction of things, but like, yeah, it takes longer, happens faster.

If you had like a three-year time span, I think we’d be more accurate, for example.

Alright.

Okay.

Well, next year we’ll look at our three-year results.

We’ll do it on you every three years now.

Anyway, thanks for listening to this podcast.

And if you guys enjoyed it, don’t forget to subscribe, like, drop us a comment, and we’ll see you in two weeks for the next one.