ChatGPT Continues to Lose GenAI Traffic Share

ChatGPT continues to lose Gemini traffic share.

Oh, this one's good. You've seen this graph for sure. Okay, so check this out. So if you look at this graph, okay, you can see this right over here. So you can see 12 months ago, the big blue one was OpenAI. That's ChatGPT. Okay. And then Gemini is red. Okay. Now if you go to today, Gemini, the red one has gotten way bigger. Okay. And then you can see these other ones here. You got Hugging Face, DeepSeek, Perplexity, all these, Right? And I'll just give you the numbers 12 months ago, okay? So ChatGPT had 87.1% share 12 months ago. Gemini was 6.4%. Today, Gemini is now 12.9%. ChatGPT is 74.1%.

Wow.

Yep. And this data is from SimilarWeb.

Most people would not agree with me on this, but I think Gemini will be one of the most popular AI platforms.

I mean, why wouldn't most people agree with it? It seems obvious to me.

Yeah, most people say ChatGPT is going to take it all. I just don't believe in that.

I think ChatGPT will be a player. I think when you look— so we should justify this, right? I think Gemini, because they own YouTube, because they own Android, because they have Chrome, because they have these TPUs, okay, because they also have Waymo as well, they have all these things, okay? They have some of the best image models. They have some of the best video models out there. They've gotten serious about how they play the game now. NotebookLM, fantastic.

Yeah, I think Google's, in my opinion, releasing better technology than OpenAI. And the most, if I had to take the last 3 to 6 months, I would say Google's doing a better job.

When I make short-form scripts, and we can talk about this, we should talk about YouTube for a second too. So I told you I've been piloting this short-form agency and they've been doing a good job. I'm like, I can see this compounding pretty significantly in the next 2, 3 months. And they write these amazing scripts and what I do with their scripts, is I take the scripts and I make a Gemini Gem. I don't pick ChatGPT like a custom GPT. The reason I pick Gemini Gem is because it can pull from YouTube. It knows what I'm like and it knows it can take from like my speaking style, my thumbnail style, and then it can take the scripts that they put together and make short form. And I've actually created a couple of good short form scripts using theirs. So I'm going to keep getting scripts. I'm going to keep working with them. But these, I picked Gemini Gem because it's hooked in with YouTube. That's why.

But from a marketing standpoint, do you use OpenAI or Google products more?

For what?

For any marketing-related tasks other than searching.

In general right now, I still use ChatGPT more.

Okay.

Because it has a lot of memory and like today, like I'll ask questions like, hey, like what's the highest leverage move here? I have these decisions to make with these 2 products over here. And then it knows because we've worked together a lot. Yeah, and then every day you get a daily pulse from ChatGPT. So, oh, but you remember that retention graph? Did you see that retention graph from earlier this week?

Uh-uh.

Oh dude, okay, so you will like this. Today's a great episode for Neil 'cause he likes data, okay? This chart over here, I don't, it took, I had to stare at this for a little bit to understand. Maybe it'll take you a little bit to get it too. Maybe you'll get it, you'll probably get it faster than me. Do you understand what's happening here?

This is, what is it, customer retention graph?

Customer retention ChatGPT, so you can see this is over the time period, number of months over time, right?

Which is actually kind of crazy 'cause that means over time people are using it more and more.

These are people who have churned, right? And they're coming back. And so the retention's going up because of memory and things like that. Yeah. This is an example. So someone said this—

Yeah, but it's not only— the interesting part here is it's not that they're just churning, they're coming back and they're not coming back at the same pace. They're coming back and starting to use it more than they did before.

Yeah.

Before they churn. And it's because of the memory.

I remember an investor tweeted, he responded to this and it's like, the only other graph I've seen like this was Uber back in the day, right? And, you know, Uber's a great company. I think ChatGPT's probably gonna be more significantly larger, but the guy that tweeted this says, this chart is so insane that it looks fake. Huge businesses, plural, will be built on top of this chart. Again, it looks like a happy face.

Yeah.

Yeah, so that just, I mean, okay, product lesson here is how do you get people to, how do you constantly think about how you can add more value and lock people in? That's what ChatGPT's doing.

But when we look at the use case for ChatGPT and even Google's AI products and Microsoft's, you know, it doesn't matter if you want to pick Copilot or Perplexity or any of these brands out there, the biggest use case is not actually to become more efficient at work. It's people using it as a new version of search, getting their questions answered. A question answered could be, what should I do with my products? Like you're talking about.

Mm-hmm.

It's just a more advanced version of search and it's where search was supposed to go. Google was just slower to adapt. Example for me recently—

But now they've caught up.

Yeah, they've caught up. Example for me, although what I don't like about Google is you don't see AI mode on an iPad. I hate that. Have you noticed that?

I don't use— besides for this podcast, I don't really use this iPad unless I'm like flying and watching movies. Okay.

I use the iPad more than I use my laptop. And when I say Like AI mode, whether it's AI mode or ChatGPT. Like the other day I was researching an accounting question and it was a really—

Of course you were.

Complex question. It was, I wouldn't say it's a complex question, but it was a complex answer I was looking for because I got hit up by someone who's a competitor. They want to sell. It's not worth my team to spend the time and energy on that company because It doesn't make enough revenue and it doesn't make enough profit, but they raised a lot of money. So then they're a C corp, I'm an S corp. So in business there's something called NOLs. Are you familiar with NOLs? Net operating losses. So typically—

Oh, you want to carry them forward?

I want to carry them forward. So I was just like, and I know they changed the laws on NOLs. So I was like, wait, let's say hypothetically they raised, you know, $30, $40 million. and I can pick up the business for $1 million. I don't care for the revenue. The revenue was small. When I say small, like $2, $3 million bucks or something like that. And when I say the profit, it was negligible, like maybe like $500,000 bucks a year in profit. So I'm like, dude, buy this company for this. This is a pain in the butt. And I was like, will the NOLs carry forward?

Mm-hmm.

So then I started doing research about the law changes and they were just like, Well, it's now based on how much you bought the business for and you can get 8% of what you bought the business for each year. So if you bought the business for, I think they were saying 8%, maybe it could have been a smaller percent, but it's something small. Like if I bought it for $100 grand, it would be more than that of course. But using easy percentages, like $8,000 a year that I can write off from my profits.

I'm like, this isn't how it's supposed to— Not worth it. Yeah. Right. But before you could.

Before you could just take the whole enchilada or not the whole thing, but a lot of it. So if someone had $50 million in venture capital and they lost it all, you had to keep in mind if a company's making profit, the profit goes against it. But let's say they have $50 million remaining, you could over time capture $50 million. The laws on this changed a while ago.

Yeah.

So then I was like, how about if I merge the companies? And they're just like, well, if you merge the companies, are you going to be the majority owner? Because if the other person, retains majority ownership, the NOL laws are different than if you retain, me being the person trying to buy it, retains majority. I'm like, well, this won't work because I would own all of it. I don't want to give them any percentage and keep those owners still around. And then I started asking questions on like, what happens if I kept it a C corp and that company miraculously grew in revenue? I'm not trying to do anything shady, but let's say we promoted it, would I get, and it started generating a lot of revenue, would I actually be able to save $50 million in losses? And they were just like, no, because you bought it for a lot less.

Yeah.

Long story short, I probably spent 15, 20 minutes when I was in my daughter's car line waiting to pick her up for school because I wait there for like more than 30 minutes. I started and I put the car in park, so I'm just asking ChatGPT all these questions. To get my answer.

But that saves you so much time and headache.

It saves me so much time and headache. That's where I found these new search tools really useful, or AI platforms, is I can do like really hardcore deep research to get my answers.

But you know what's important too that you're mentioning? It makes you a better strategist, but you have to have that level of thinking already. That experience matters a lot.

Dude, you're totally right, because I've asked it a lot of accounting questions, like Even, I have one corporation that used to make a lot of money and I've merged it in with other corporations. It was an S. The way we did the assets, the company that was an S, it ended up turning into a C and I don't own any of the assets anymore, but the C is, think of it like an empty shell.

Uh-huh.

So it doesn't generate any revenue, but I still keep the company around because I like it and You know, after the 5-year limit, I can convert it back into the S and I'll, you know, I'll do whatever I did with that corporation. So with that company, you know, I may have to end up refunding someone, which is fine. And I already predicted this and I was like, okay, if it's a C corporation now, you know, if I can end up, if I refund, it actually has negative income, But the income generated when I was an S, so in theory I should be able to claim that off my personal taxes. And they were like, yes, you can, as long as you have enough income within the year to offset it. And I don't know what I'm going to generate each year because sometimes we reinvest a lot, we buy more businesses, et cetera. And I was like, what happens if I don't? And they're just like, you'll end up— it goes away because it's too late to go redo your taxes because it's been more than 3 years. So I was just like, so whatever you save in that year is what you save. The remainder balance that you refunded, you don't get a tax break on it. And I was like, I'm like, you're asking me if I do a credit. What happens if I turn it into NOL? Because I should be able to. And they're like, actually, you are correct. You got the, you know, they're like, this is something that, you know, that AI did not think about. And I was like, what about turning into NOL? And it was like, yes, you can turn it into NOL. Use it as a debit instead of a credit, and you can wipe off all the gains on your personal. And I was like, cool, sounds good. And I was like, if I should do this, because when I had that money and I paid taxable income, I'm not trying to do anything shady here. I lived in a state where there was no state income tax. Now I live in a state where there's a state income. And I was like, because it's an NOL, in theory, I should be able to save on state income tax, even though I didn't pay state income tax on the original amount. And they're like, that is correct. You will get a rebate on state income tax as well. So I was like, oh, cool.

Yeah.

But the AI did not think through all the different scenarios and I had to feed it data on, or I had to give it ideas on how it could work for what I was looking for.

Let me tell you why this is important. So again, this is new work that couldn't be created before because Neil's old behavior is he might, you know, call a handful of different people all the time. He used to call people all the time, right? Now he will pester the AI instead, and once it's nailed down, then he will make the call to double-check with a professional.

Yeah, you're so right, because normally I would have called 3 5 accountants, maybe 2 of them would've picked up and I would've spitballed with them to try to figure this out. And then once we would've had a plan, I would've verified it with another accountant to make sure it's legal and it passes and I'm not doing anything shady. Now I pester the AI for anything I want.

Yep.

And then I verify it with a human who's an expert. And I'll tell you this, in most scenarios, I'm using AI for really high-leverage stuff, like I'm talking about millions of dollars in taxes here, right? I consider this high leverage. This is stuff I'm figuring out in less than 30 minutes, so it's very valuable. I could have figured it out by talking through quite a few people, but it's easier now for me to go through AI and then verify it with human. But if you yourself as an individual aren't decently knowledgeable in the subject, and ideally more than decently knowledgeable, like I think I know taxes and accounting pretty well, even though that's not my specialty.

You definitely do.

The AI doesn't produce what you're actually looking for. It is not that useful. You need the person who's prompting to have the expertise themselves, or it's not that useful. And this is for any subject from what we really see. Like if you're in marketing, if the person behind there is more talented, they're going to have much more success using AI. And the second thing is you should verify it by an expert. And Eric nailed this and he knows me really well. And why would you say that I'm going to ask an expert to verify this?

Well, because you need to make sure it's— because that's not an expert first and foremost. You need to make sure you're legally okay. And you probably will ask what you probably asked 2 or 3 actually.

Correct.

Yeah.

Because you don't want to go wrong. And Eric's spot on. I've had a lot of times where I've asked experts after doing my research and OpenAI is wrong or ChatGPT is wrong with what they gave me.

How often is it wrong though?

For this complex stuff?

20, 30% of the time?

More.

40%?

Yeah, it's right more than half the time, but probably 40% of the time it is.

But it still moves you a little further than you would before. Because the thing is, once you call all these people, you exhaust them a little bit. When you do it in ChatGPT, even if you get a little closer, like a mile or 2 further, it's still worth it.

Yes.

Yeah.

Because even when I talk to accountants on some of my cockamamie ideas, Sometimes they have to do research too.

Yeah. Yeah. See, this is where the AI is good. It's infatigable.

Yeah. Because when I say my cockamamie ideas, I'm not trying to do anything illegal. I just have random ideas and it's either, it's pretty black and white. Either it'll work or it won't legally.

Either it will make money or it won't. Exactly. That's how your mind works.

And the problem with a lot of the stuff that I want to research, A lot of them are about different laws and rules and cultures in different countries because it's not all accounting.

Someone has to do it in a bunch of spaces.

I have it do it for different countries and then I go like, let's say if I was talking about accounting, I would go do the research in ChatGPT for multiple countries and then I'd bring on our international tax team as well as our domestic tax team because our US entity owns all the international. And then I'm making sure everyone's on the same page based on the research. But it's really tough to do that with one individual because one individual Doesn't know every single country's tax laws.

Dude, you need to take all the transcripts of you working with this thing and make like sub-agents for each region and then have it work with your accounts in each region. And it's just working on that the whole time.

That would be more efficient than what I'm doing.

That's what— but you will do that eventually. That's where we're going, guys. So, by the way, I think the takeaway for this one, and we can move on, is we use it for the highest leverage things. So for me, I'll workshop with it like, okay, what's the best comp package I can have for this executive over here? Or Sometimes I'll even like, because I work with it so much, like, hey, I'm like, hey, we'll workshop something. And it's like, okay, here's the new mission statement for the company. Here's what we should be focusing on over here. And then I'll often ask it to ask it like, hey, of all the things we worked on this year, how high leverage is this? Okay, what do I need to reprioritize? And it helps me with that because for me, I have the entrepreneurial ADD. I like to work on a lot of different things, but this keeps me focused actually. And that makes it so worth it. Yes.

And where I love using AI in marketing from a global standpoint is Do you remember like the ads of like Coke and Pepsi? Like it'll be like a can, like dressed as a Pepsi can in a Halloween.

Oh yeah, yeah, yeah, yeah, yeah.

Or a FedEx truck that's kind of disguised as a UPS truck and they're making fun of the competitor.

Not all heroes wear capes or whatever that was. Yeah, yeah.

That was the Pepsi one.

Yeah.

We were running a campaign in Japan and someone wanted to run a competitive campaign, and I knew the company is a really big company, so I started asking. you know, I was going to ask my team in Japan, but I was like, dude, my team is sleeping right now. I don't want to wait for them to get up because Japan and the time zone versus me and PST. I started asking ChatGPT, you know, like, how will this work, this ad work in Japan? Because they wanted to do a Japanese version of it. And this was a global company, but I was talking with people based out of the US. They're like, yeah, we should just push it because global marketing was run here and they would just tell people what they want done and people adjust things to local markets. I'm like, I don't know if it'll work. So I was spitballing with ChatGPT for probably like 20 to 30 minutes to try to figure out how to make this work. And I know in Japan people are very loyal to brands.

Yeah.

They don't just switch.

Yep.

And what I ended up finding out by doing ChatGPT, and I had a hunch, but I'm not certain because I'm not Japanese, they ended up saying there's a good chance you're going to offend people and they wouldn't like it. So then we decided, we did confirm with people in Japan, but ChatGPT gave us a quick answer and I was able to call the company back on Pacific Standard Time. I'm like, I don't think it's going to work. My gut inclination is people are loyal to brands and they won't like it. And I'm like, this is what ChatGPT is saying. I was like, I think we shouldn't spend too much time on this until we get feedback from the regional people.

Yep.

I shot off an email.

Yep.

And I didn't shot off an email with the points that ChatGPT was breaking down, but I just wanted to see what the regional people thought. And I'm like, just be blunt. If you love it, let us know. If you think it won't work, tell us why in just one sentence, because I don't want people wasting too much time. And no joke, our team and their team told us back in one sentence why it wouldn't, and they both said it's going to offend people.

Yeah. Okay, before we move on here, like, by the way, I also like using ChatGPT to settle arguments too. And the majority of the time I'm right, thank God. But it's so good because I then I'll just like, sometimes I'll be walking with my dad on the weekend, I'll just open up audio mode, right? And then whatever crazy stuff he might say, I just like do it. And then like, even then with an AI, he won't change his mind, right? Which is insane to me. But I'll even do it with like, you know, other situations too. We'll leave it at that. But like people do change when you take, your emotions out of it and it's just a robot talking, they have to, they usually have to concede. If they don't, maybe there's something off. But yeah.

I would say, and you and I have used that as well when we're dealing with each other, I would say it works really well for black and white type of things. For stuff that's ambiguous in gray areas, it's tough.

Religion or politics, it's tougher.

Or stuff that there's not a ton of data and it's just giving opinion, it's tough as well. Yeah.

If you're in B2B, here's the deal. LinkedIn account-based marketing sucks, but It doesn't have to. With Karrot, it allows you to create personalized ads towards your customers and also personalized landing pages that align with the ads. So you're going to have higher conversions and influence pipeline reporting to prove that your ABM works. So if you want to learn more, just go to karrot.ai and you can join the waitlist. Again, it's Karrot, K-A-R-R-O-T.ai, and we'll see you on the other side. Did you know that many of the best companies go slow before exploding their revenue? So I'm going to give some examples over here.

Many companies go slow?

Yeah, before exploding their revenue. So when you think about, so ChartMogul here, okay, Lovable spent 18 months building before their breakthrough success. So Lovable I think is doing 9 figures a year now. Clay grinded, you know Clay, right? Clay's the clay.com, it's like the sales thing, right?

But what's Clay Grinder?

Sorry, they grinded for a very long time. Yeah, so basically Kyle Poyar tweeted this. So basically only 3.3% of SaaS startups reach $1 million ARR in less than a year, and just 1 in 1,000 reach $10 million in less than a year, right? And so here you have Lovable, it's been 18 months building. Clay grinded for 6 years before seeing revenue traction, meaning that they had to sit it out for a long time. Bolt.News, I mean, they spent 7 years crawling their way to $700K in ARR before their new product exploded, which is, I think now they're doing like close to 9 figures or something like that. But the point is, a lot of these situations, it just takes a long time to grow. And Replit, I think, took over what, like 8, 9, 10 years or whatever. And now they're doing, you know, multiple 9 figures. Um, so many of the best companies, they go slow before exploding their revenue.

Yeah, but people believe everything will happen overnight. And you and I know that's not the scenario. Yeah.

They believe it's an AI growth. They believe it's like young kids too.

Yeah, it's not.

Most founders are like 40+ years old, the successful, like big companies, right?

Really? I didn't know it was that old. I assumed it was in the early 30s.

It's like 45, 50.

Oh, okay.

Yeah. Well, it's 30s because we've seen Mark Zuckerberg and we've seen like these people.

Yeah. Matt Mullenweg and a few of those.

Oh, interesting.

Yeah. He was young too, right? When he started Automattic. I think that was his only company he created.

I could be wrong, but yeah. Anyway, guys, that's it for today. Please don't forget to rate, view, subscribe, and we'll see you tomorrow.