Traffic from AI Chat is Growing Faster Than We Thought

My first reaction was, oh, this is gonna take years to pay off. And then you showed me Scrunch and I started thinking about it more and we have the stats from some of the people who work with Hypergrowth Partners. I was like, oh no, actually this is vital right now. Okay, two big topics today and a smaller one. We're gonna first talk about AI chatbot traffic and a database fight and how to deal with trolls, and then maybe some other stuff about fun conference stuff that I'm doing. So with the AI chatbot traffic, our friend Kevin Indig, who has a great newsletter called The Growth Memo, he always talks about SEO. It's probably the only thing I read on SEO, and he got a hold of a bunch of data and this study on traffic from GPT effectively and where does that traffic go. It also had traffic from Perplexity and other stuff too, but you'll have to get into it. The main things are it's still small as a percentage of all traffic, but it's growing quickly. It's mostly sending people to e-commerce sites right now, the usual suspects Amazon, Target, etc. But the users that it sends over to those sites, they're engaged, they spend more time on the site, they go deeper. And also interestingly, which might seem counterfactual to that, is it's mostly sending them to home pages. And so those are a few of the takeaways. He had a few more, you should check it out. But we've talked about if you should be doing more to get the SEO of chat. I wonder what term will come up with that, it'll be another three-letter acronym. But yeah, Gonto, what are your thoughts?

Yeah, so first of all, like a lot, we've seen this happen in a lot of Hypergrowth Partners customers. I know that for neon.tech, for Aptakube, and for Clerk, ChatGPT and other similar ones we're starting to grow and grow as one of the sources. For Aptakube, for example, it was in the top three sources, which is insane. So at least like we're starting to see this in B2B happening and not just for the e-commerce, which I think is fascinating. Other than that, I actually work, one of the companies I advise is called Scrunch AI, and what they do is basically help you understand how people use ChatGPT and other AI platforms and tell you like where are your ads, how it works, etc. They have some competitors, but main takeaways that I take for, first of all, like check Scrunch AI, maybe it's a good acquisition. I'm actually making an intro to Hank today. But other than that, my main takeaways from it that I think is fascinating is number one, being on top on, are you showing up on prompts in ChatGPT and similar, I think is really, really important. Number two is if you show up, it's very important in which position you're at. Like are you in the top, in the middle, or in the bottom? And what is the sentiment that the chat has about you? Because sometimes it mentions you but for the worse. So looking at multiple prompts for that I think is, it's the base things. Two things that I think are very important to check is number one is, before, with typical SEO with Google, you have this concept of like pages that were linking to you, like link building. Using ChatGPT has something similar where ChatGPT and Claude and Gemini will use search and they will search pages to then give you a result. So what you could do potentially is if you, I don't know, checking 500 prompts, which again is a bit of what this Scrunch does, it can actually tell you which pages are being searched and used in some of the prompts. So it will take like which page is used in how many prompts and is it a third party page or from you. So if it's from you, you can improve it. But if it's third party, you can potentially partner with them or work with them to show up differently or something like that. But understanding what are sources that ChatGPT uses and actually connecting with them I think is really, really, really important. Last thing that I'll mention, ChatGPT has memory, this concept where it remembers who you are and what you do. A lot of people mimic this with Claude and with others by having a custom instruction on who they are. What's fascinating about this is that replies and searches that ChatGPT or Claude will do will change depending on whether you say I work as a marketer or a salesperson or whatever.

So actually understanding who is your target, and maybe you have two personas, maybe you care about a marketer and a salesperson, and seeing how ChatGPT searches for salespeople and for marketing and how you're mentioned differently in each and which results from search are used, I also think is very important. Because it's not like there's only one, and you should think about which exactly is that persona. Those are the main things I think you should care about. There's Scrunch, there's Profound, there's other services that are doing this for you, but it's something I think you should watch out for.

Yeah, my first reaction was, oh, this is gonna take years to pay off any effort into this. Which is funny because, you know, as we've talked about on the podcast, one of my themes for this year is to get really good at using AI at work. And then you showed me Scrunch and I started thinking about it more, and we have the stats from some of the people who work with Hypergrowth Partners, and it's like, oh no, actually this is vital right now. And I wonder how self-reinforcing it becomes. I wonder if it's important to get in early or you'll get left behind, or how often these things will kind of reset on their recommendations to you. But I also wanted to double click on what you said about specifying the audience and the persona. I feel like it's more important than ever to be clear in your messaging who is your product for. Because that informs, GPT knows I'm a marketer, it knows I'm a marketer at Laravel, that I do stuff with Hypergrowth, that I have a podcast. It knows so much about me and I always see, sometimes I'll tell it something new about me and it goes memory updated. And like, oh, I don't, you know, I don't know how I feel about it. But then it becomes more useful to me and its recommendations become stronger. And I think everybody needs to recognize that these things are just going to take over, the traffic is going to continue to accelerate. So I don't know, we got to figure out. It won't take years to pay off, it could take months to pay off.

And Google is now testing changing search results to an AI overview for everybody. I saw that last week. So if Google ends up doing something like that by default, like this would be true right now for everything. But again, and I think we talked about this in another episode, but it's still important to understand that this doesn't change SEO fully because all of these tools still search either in Google or in Bing or in something. And the main difference I think is that it doesn't matter if you're first result now, it matters if you're in the top 10 and actually being mentioned in multiple of them. So more important to be mentioned in multiple of those results rather than just the first one and show up like it used to be, which is something we should think about. And similarly, a lot of these tools use Bing now, so Bing is starting to become really important. It's not just Google.

Yeah, Kevin talked about that in his thing, the forever second search engine of Bing actually getting market share. Exactly, it's very interesting, important to pay attention to. You know, I don't have any other takeaways than that. It was just really interesting content.

I agree, and you should also subscribe to Growth Memo from Kevin Indig on Substack. It's fantastic. I think it's the best SEO resource out there, and I say that not only because he's one of our partners at Hypergrowth, but actually because I believe it as well. With that, let's move to database fights. There's been some messages from Dax on Twitter as well as some messages from Average Database that were, actually that is basically mostly a troll that is anonymous but that owns a database. We were discussing whether it's the CTO or if it's somebody else, we don't know. But there was a big fight here where Average Database and Dax were publishing some stats comparing PlanetScale to neon.tech to Prisma database. And in that specific chart, what he was saying is that PlanetScale is incredible and that Prisma and Neon are sort of similar, but Prisma ends up being better. Nikita actually answered to that saying that that's not true and that basically the test, how it's done, it triggers a cold start from Neon that in some cases, it's not, in most cases it's not needed. And he was saying how Prisma doesn't have that tech. And then that tweet got a bit of sheds from both Average Database as well as from others. But what we wanted to chat a bit about here is you will always get trolls that are anonymous or not like Average Database, or you will always get posts from people that are smart like Dax that are well intentioned but maybe are not entirely correct. So if you are Neon, how would you think, how would you react to this? If you were to do it next time, or how would you deal with these trolls and some of these posts that maybe are not intentional but hurt your brand? Like, what's your take, Hank?

It's really hard. So first, with an obvious troll like the Average Database CEO Twitter account, it's a troll account. It's trolled me before even, and I've muted him, I should just block him. They told me on a funny post I wrote about how I bribed a Vercel employee with golf clubs to get something done, and they just, they couldn't help themselves. So what was interesting is they have an ulterior motive, a clear ulterior motive. It seems like, okay, they're selling a database as well. Anything that makes other databases look bad and gives them a chance to get a further wedge, like of course. And the thing is, with any traction they get, there's gonna be some base of that traction is already their fans. And so anything you do to fight that, you're fighting against people who are already fans, and you're just gonna roll around in the mud with them. So my general thesis on people and accounts like this is why even engage? It's better to just let it be. I'm curious if you agree with that, and then I think we talk about like dealing with a Dax.

I don't actually. Like, I do agree that you shouldn't engage directly because that's exactly what they want. It's an engagement trap. Like if you reply, as you said, there's gonna be fans, they're gonna get shared and stuff like that. But I think there's a way to reply that is not replying, where you reply by posting something. Like, for example, in this case what I probably would have done is, I'm seeing this post from Average Database saying like, oh, like Neon is so fucking slow, Prisma is much better and even more PlanetScale or something like that. And if the problem is that the benchmark is done incorrectly or not in a way that, actually in a way that hurts Neon, what I would do is actually build an open source performance checker that actually checks performance in a better way. Because the problem here was that this tester was just a ping from a Cloudflare Worker that is actually further away from all the devices that any human would be, so it's not even accurate. So what I would do in this case is do something like a benchmark that is open source, and then I would post not only the results but also actually post that it's open source and asking Prisma and PlanetScale, or like, hey, this is open source, we try to be fair for everybody, but if you feel we weren't, send a pull request, we can review it and include your stuff in here. So then I think on one side you look better because you're doing something open source and you're open and you're doing a real performance test. And on the other side, you're leaving the door open for the competitors to actually do or include something about it. So it is a response in some way, but it's not in another, because it's not like, oh, what the fuck, like the competitor is wrong or the test is wrong. You actually build it, everyone, you make it open source, you make it inclusive of everybody, and you ship that as an answer. That's, that's I think what I would do.

Yeah, interesting. Okay, so you're saying you still don't reply to the troll.

Directly. Instead you come up with a new thing. Because Nikita did reply that, hey, this test is designed to trigger Neon cold starts. Which he also adds that they have a feature that this thing is tripping over that Prisma doesn't even have, so Prisma couldn't even, like, exactly that. And you know, it has something to do with hibernation or scaling or something. So yeah, what he could have done, and he did it as a reply in the thread so it got the least amount of attention, but the problem is not

that it was a reply in the thread, but rather that not only did he defend Neon, but he talked bad about Prisma. You do not, first, for like, I don't think you reply, but second, you do not talk about the competition.

I would agree with that as well. So a good way to go about it would be wait a day. Like take a day, chill, let the thread do whatever it does, and then you could just start a new thread that says, hey, here's how to properly benchmark databases. We've open sourced this test, you know, Prisma, Supabase, whoever, you can contribute whatever you like to this benchmark. You could create different scenarios, whatever, but here's how you should actually properly benchmark it. And then the people will, you know, perhaps reply. And you could even reply to one of those threads to maybe someone who's a less clear troll, like the Dax thread, and say, hey, here's our response, like we think there should be an open test or whatever. I like that. Exactly. Sorry,

I interrupted, but with Dax I would even go one step further, and because he's not a troll, I would send a DM. So not only I would do that, but actually send a DM to talk to him because he's a reasonable individual, I think.

Yeah, well, and Dax in the thread did correct some people on some stuff, and he opened his thread with it's a stupid test.

Yes. And he also called out why it's a stupid test in more detail in the thread. I think it's still, overall, the fact that this stuff is spreading is going to be damaging. I think unless you're PlanetScale in the middle with your little green squares, and that's kind of the point of some of these. I don't know, benchmarks are always scary and hard in these very commoditized businesses. And yeah, I guess there's another topic here of benchmarks and how you should be thinking about them, and how, you know, there's a world where Neon and the rest, where they already have proactive benchmark testing that they're, you know, publishing or open sourcing somewhere. But that really comes down to like a belief in your product and how you are open about it or whatever, and that's hard to do. I think in

some places benchmarks don't make sense. Like I don't know if your product is more about what's the developer experience and how good it is and blah blah blah blah, then it doesn't make sense. But in these cases where you could use PlanetScale, Neon, Prisma, whatever with any of the existing frameworks and just put the connection string because they serve you like a serverless database, I think a benchmark can really help. Because they can have some features on top, which all of them do, that then target different folks. But in the core it's still a database, and I think latency and benchmarks matter so much more in these type of businesses than in others.

Last topic. It's a weird one because it's not really a topic, but like, we talked before the episode with Hank and I think Hank did some really cool unique stuff on the Laravel conference that he organized. He tries not to talk about Laravel, but I told him that he should bring this up because these two things, especially the early bird one that he'll comment, I think were really good. So I'll just let you talk on this too.

Yeah, don't want this to be a Laravel marketing podcast, but there are two fun things we're doing conference wise. So one, I was at Laracon EU in Amsterdam last week and Gonto said it's a shame I don't do drugs because that's the place to be for that. But I don't know if you want to keep that in there, I'm totally

Okay with the drugs, there

you go. You can keep it in. So one thing we did, okay, so we started selling Laracon US tickets last week, and an interesting thing about early bird tickets is they're the most discounted but they go to the people who want to go the most. And the economist in me sees consumer surplus and thinks, what can I do with that? We're not interested in making more money off the event. We run Laracon US as a break-even event, but I figured, well, if I really want to go to the event and I'm willing to spend more, what would I want if I spent more? So we came up with a community bundle ticket, and it's an adaptation because we used to have a community sponsor where you pay a thousand bucks and you can just put your name up and you get one ticket. And I thought, well, let's make that just a ticket for anybody, not just companies. We'll let them put a name or brand up, whatever, but then we're also going to give them a swag box. We're getting them a cool jacket, the Artisan jacket, because Laravel's for web artisans. And we're also going to give them express check-in and coffee with the Laravel team. So we're giving them a bunch of cool swag and a cool experience that's a little more exclusive. And so far like 10 or 15 percent of the tickets sold are those type, so we're gonna have to cap it really soon. We're almost out of those because I can't, I can't let everybody have coffee with the team like that. It'll be madness, then it'll just be coffee with a bunch of other people who paid instead of coffee with the team.

Before you go to the next one, what I thought was really good about this is that I never thought about it, but you're right where every conference does this, that they have this early bird and the people who buy the early bird are the people who love you the most and are probably more willing to pay. And what got me thinking about that is that there's probably other places, not just a conference, where you give things cheaper or easier or stuff like that to your lovers, and then you give harder or more expensive things to the people who are more on the fence. And it shouldn't be the opposite. So just stop and think about, is this happening in other parts of my company, and could I do something different?

Yeah, and you know, to not be gross about it, what you can't be in an open source community like Laravel, like I can't just charge them more because they're willing to pay more. You know, the thought was, can you give them something special? Can you make them an even bigger, more loyal fan by giving them something awesome that nobody else will get? So that's the idea here. Glad you liked it, and I think there are lots of areas for this type of thinking in B2B marketing. Similar thing, we've talked about like enterprise marketing dinners before. We did one at Laracon EU last week, and one thing I tried that's different than anything I've ever tried was, normally in marketing we like to be subtle, and this time we did the opposite of subtle. I had my CEO tweet out, if you're at Laracon EU and you spend $30,000 plus a year on AWS, then DM Hank Taylor and we'll take you to dinner and show you some cool stuff with Laravel Cloud, our upcoming product. So instead of the normal shy, like, hey, we think you're special, come to dinner, we were just like, hey, if you're here and you have money, let us take you to dinner, you're our ICP. And it was the strangest reaction. I suddenly had a bunch of people DMing me screenshots of their AWS bill wanting to come to this dinner. And it was an awesome dinner. A lot of people said it was like the coolest experience out of Laracon that they've had, like being able to mingle with the team and do some cool stuff. It wasn't even really a pitch, it was mostly we just chatted to people. And it went really well. We generated a bunch of enterprise pipeline, which we don't have an enterprise product yet, but we're gearing up to it, so we wanted to test interest and get requirements and that sort of stuff. So it was cool, and I told you about it, you sounded like you liked it.

Yeah, I liked it. And the one reflection that I have, all these, it's actually not regarding work, it's about personal life. But it's like, the more money that people have, the more things they can get for free. That's how unfair our life is. Like I see it all the time, like you're Messi, the best football player in the world, and you can go to any restaurant for free. You can go to the Super Bowl, he was invited for free in the best seat. It's just mind-blowing, the more you have, the more you don't need it.

Yeah, you're into like the three levels of wealth stuff right now, aren't you? Aren't there some memes about that?

I love it. My favorite one of course was the LinkedIn one. It was like the no LinkedIn, LinkedIn, no LinkedIn. Like I want to get to that. I haven't gotten to a no LinkedIn yet. I hope I eventually do. Excellent.

Well, we can all aspire. All right, I gotta jump, you gotta jump. Good pod. Thank you everybody. Cheers.