Webit 2026: AI for attention: how the world's biggest creators engineer retention - Mario Joos

This talk explores the significance of AI in the attention economy, particularly how it is used to enhance viewer retention for digital content creators. The speaker, Mario Juice, discusses his experience in optimizing video engagement for major creators like Mr. Beast. He highlights the importance of capturing and maintaining audience attention through various strategies, including the use of compelling characters, relevant topics, and innovative content formats. Juice emphasizes the role of AI tools in streamlining the creative process, from analyzing viewer engagement patterns to optimizing thumbnails. He also addresses the challenges of content production costs and the potential of AI to democratize access to these tools, allowing more creators to succeed. Ultimately, the session underscores the balance between leveraging AI and preserving the creative essence of content.

Now we enter one of the most dynamic arenas of the AI economy, the attention economy. For centuries, businesses competed for resources. Today, they compete for something far more scarce, human attention. In a world of infinite content, infinite choice, and shrinking attention spans, the ability to capture and retain attention has become one of the most valuable competitive advantages on the planet. And few people understand this better than

our next speaker. Working behind some of the world's largest creator ecosystems, he helps build the systems that keep hundreds of millions of people engaged, returning, and connected. Because what begins in the creator economy rarely stays there for long. The lessons learned by the world's most successful creators often become the playbook for brands, media companies, retailers, and businesses everywhere. On stage, we welcome Mario Juice, retention director for

creators with more than 100 million subscribers, including Mr. Beast, Stokes Twins, and Allen's Universe. AI for attention. How the world's biggest creators engineer retention. Thank you. Thank you. Surprise. We're going to talk about AI. Um I'm going to talk to you about attention. Um thank you for the introduction as well. Uh I had an introduction slide as well, but um I think we're going to be able

to skip over that one. So for the last six years, I've been focused on attention or what in the YouTube game we call this retention. and uh been helping some of the biggest creators how to make people watch for longer, how to grab their attention, and how to make sure that people feel good about giving you that attention. And uh a lot of those videos have scored

over hundreds of millions of views. And the goal of this presentation is to give you an insight in both what we previously did for that attention and then also help you get an insight in how we now use AI for that same attention to help speed up the process. So I want to show you something first. I'm not sure if you've maybe you've dealt with Instagram, maybe

you've dealt with Facebook or in this case YouTube. You might have seen a chart like this. This is mainly for video content. This chart, if you recognize it, is what we call a retention chart. And on this chart, what you're going to be able to see is exactly the moments that people keep on watching or that the moment that people tune out. Now, sometimes these charts give

us second by second information about how the audience interacts with your content. And sometimes, well, sometimes it's you have a data point every few seconds. But regardless of how accurate this data is, when you analyze thousands of these charts, you get a lot of insights of what people like, what they don't like, um how they interact, which type of characters, which type of people they like. U

so much stuff. I mean, we'll soon learn how much there is behind this chart. So, because of that one, we want to optimize for that one to reach a wider audience. So, this is a quick introduction slide as well. Um, my name is Mario. I am a I have a small little company. It's a research consulting company, but I would say it's more like a think tank

where we put a lot of bright minds together and we think about these these topics such as uh attention. We also did the research on attention spans. And to give you an idea, attention spans are not actually shortening, but that's a whole other conversation. Um, so we've done over 50 billion views with most of these creators. Some creators still score over a billion every single month. Mr.

Beast, etc. So those are the type of clients that we typically work with. Now I want to give one disclaimer before I actually start revealing a lot of information and this is a disclaimer about AI. So this little keynote that I have here is going to sound very much like I'm very very for AI and I really believe that AI can elevate us, can support us, but

there's a lot of dangers especially in the name of creativity. For example, actually as I'm standing on the stage, part of my team right now is in the EU Parliament talking about AI policy for the sake of protecting creativity. So I don't want this to seem like it's just like, hey, let's have everything replaced by AI. AI supposeds to support us. And I really hope that in

my presentation that that's also very clear. So that being said, when it comes down to attention, there are three big factors, three key key components that contribute to attention and that you have to optimize for whether that's organic content, paid content, doesn't matter. Especially with organic content, actually this is very important because the algorithms are optimized to recognize how well you do in all these three categories.

So these three categories are we have to be good at capturing attention. This is mostly for people who've never interacted with your videos before, but sometimes even the returning customer or the returning viewer. We still want to make sure they stay interested in our stuff. The second component here is that we want to maintain the attention. As algorithms now are designed, there's this formula called expected watch

time per impression. Watch time tells the algorithm like, hey, this viewer really likes this content. It's one way of measuring satisfaction. Whether it's a flawed way or not, that doesn't matter. There are some flaws. That's why I'm mentioning this one. But when somebody is willing to give you 10 minutes of their time, they're probably enjoying it versus somebody who swipes away after 4 seconds. And last thing

is we want to make sure that every single time that somebody interacts with your content that they feel like it was worth the experience. I want you to think about a restaurant. When you went to a restaurant and you everything sounds really good. It's a premium pizza restaurant. You ate the pizza. It tastes bad. Are you ever coming back? Not really. So that's the same with videos.

If people feel like ah nah they're not going to come back and next time they won't even trust you no matter how well you sell it. So on those things I have here two examples. It's uh on top you have a example of Mr. Beast. It was a Squid Game video. I've worked on this one as well. And the second one is from a creator called Law

by Mike. He's a lawyer who makes edutainment content. In this case we had a video I spent 24 hours with real criminals that you would actually spend time and actually see how would they break into a house. So I want to talk about these three categories and talk about before how does all this judgment work and what are the kind of things that we have to think

about for grabbing attention. So in the case of this I'm going to put all three on screen. We have in both scenarios we have a high stakes environment. What we've learned is that if we create a video idea if we just put it in a high stakes environment that is more likely to grab attention versus when we don't put it in a high stakes environment. So especially

with these really big creators we often try to create that one. Now, in the second case, we have relevant topics. Now, Squid Game at that time was very relevant. I'm quite sure that I'd be surprised if nobody has either seen or heard of Squid Game before. Uh, the only thing that has ever come as close to Squid Game is probably now the recent K-pop demon hunters. If

you got young kids, you probably know what that is. So, but in the second case, a lot of people in the world have an interest in criminals and especially this is why crime shows are so popular. So we knew that criminals is better than for example spending 24 hours with my pizzeria delivery guy for example. And then novelty is like every idea becomes more interesting especially if

it's an idea that you're familiar with once there's a twist. Squid game but then it's in real life criminals but we're actually spending 24 hours with them. That becomes very very interesting. So on that topic that's what helps us grab attention. Those are the factors that we think about. Now, these two screenshots that you will see here are from those same videos. Grabbing attention isn't enough. The

most important is to maintain that attention. And there is three things I want to talk about. Let's go one slide back. There you go. There's three things I want to talk about with these videos. So, it's important to have interesting and unique characters because nobody likes to watch boring people. Uh, and there are actually techniques and things that we look for. For example, we know that if

somebody has high status, they're naturally more interesting than a lower status person. If somebody has extraordinary skills or abilities, that's actually more interesting. Or if somebody is just weird, you know, Jack Black type of weird, we find them more interesting. So, if we don't have those things, people aren't interesting. So, here you have the example, we have interesting characters because of who they are. Um, in the

upper case, they were roommates and they betrayed each other. that's like uh roommates that betray each other at that level was more interesting. And on the bottom you have like this very sticky lawyer and then we have a real criminal. You don't see it very well but he has tattoos all over his face and stuff. It's very very interesting. So we do that as well. Then also

in both cases we break the fourth wall. We have the camera uh on the top. It's very difficult to see from here but she is wearing a GoPro on her chest and we see it in the frame here. We see that she's wearing the GoPro. So, we're we're making it seem like, okay, this is a production. It is not like a movie. And last thing, what you

cannot hear right now is that both of these videos, these were tension scenes. So, we have a lot of music involved. Music is one of the best ways to keep attention. Now, I will say if you make ads, for example, and you put this generic corporate music that you found on one of those websites, I can tell you, you will lose attention so fast. Start replacing that

music with tension music. Even if it doesn't feel like it fits, I can guarantee you people pay more attention. And the last thing that we do and that we spend a lot of time on is rewarding that attention. In the top here, we have that ending that really feels earned and deserved. And in the bottom, what we have is during the video, we actually make it into

the house. So, whatever we promise, we make sure there's always a reward. This is a big issue with, for example, movie trailers. They don't often reward your attention. They just tease and tease and tease. No, even in a movie trailer and I teach this to people who make movies like your job is to already give things in your trailer so that the viewer gets a taste of

what's good. And then overall our brain always judges was this worth my time or not. So a lot of small things I don't want to go too deep into this because that's not the goal of this presentation. The goal of this presentation is to highlight the problem that we have with all this. And that is that those features that I just showed you, they have a cost

between $20,000 and $4 million every single time. And these are not companies that can generate that amount of money. I mean, Mr. Beast at this point, yes, these are creators that often rely on AdSense, which per thousand views, you get about $3 to $5. That's not great, you know. So, obviously very expensive and inaccessible to 99% of creators who are ambitious. And that's where AI really provides

us with a solution. So over the next few slides, what I want to do with you is I want to show you the tools that we've been using to both save ourselves sometimes some time uh save ourselves money and just sometimes speed up the tedious processes. Like I like to think big picture. I don't want to be thinking anymore about oh you know should I oh there's

corporate music here. Maybe I should replace that. If we can use AI for that would be fantastic. So, here's the three tools for grabbing attention. And um this is a screenshot. I'll describe what you see on the screenshot. In case you wear glasses like me, I wouldn't be able to read this. But basically, what we can now do is if we have a thumbnail and a title,

we can run this through tools and detect a lot of these minuscule factors. What's the composition of the thumbnail? How do the elements work together? What is more likely to grab attention? And there are tools specifically designed. This is not just running something through CHP and relying on a generic info source. This is literally using those tools to then figure out exactly, hey, how can I optimize

this one? Now, these tools, they help us with a lot of things. They can help us create new thumbnails. They can help us create find potential flaws. AB testing, testing multiple variants at a rapid speed without paying $500 per thumbnail. They can help us recreate similar thumbnails for other videos. and so on and so on. These are all customtrained tools specifically for this purpose and that's how

we use them. I really recommend you to do the same. Now, this is an example of a recreation. There's a thumbnail and we're like, "Oh, we want to test multiple variants, but small variations." Run it through this tool. Bam. Different variations look exactly the same. This just saved us $1,000 right there because I didn't need to hire a expensive graphic designer anymore. Next up is maintaining attention.

This is my expertise. There are thousands of factors that affect when somebody's going to tune out or not. These are tools designed that you run a script through and it will detect you anything. Potential uninteresting characters, repetition, uh story lines that go nowhere, unresolved conflict, all those kind of things can easily be detected. You don't need a top expert. The issue is, and the reason why these

tools are so good, an actual expert, and I'm being very transparent about this, can easily cost the top experts easily cost you 10 to 20K per month, non-exclusive. Wild. That's not accessible for somebody who makes,000 bucks for like a video with 200,000 views. So, if you have AI tools that can help you do 80% of work, absolutely amazing. Now, uh, yeah, detecting patterns already. The rest of

the slide I'm going to go through very quickly because I see that I'm slowly running out of time. Now, one disclaimer about everything that I've shown. This is, by the way, if you haven't focused on this yet, this is probably your number one thing you should focus on. AI currently utilizes what we call explicit knowledge. This is knowledge that is documented knowledge that we know. This doesn't

include intuition, you know, etc. And all these tools, they were great, but why are experts still alive? Because AI doesn't have something what we call tacet knowledge. A lot of the essential knowledge for businesses, for even creators, a lot of that essential knowledge is in people's their minds and hasn't been documented yet. So that's the reason why always you don't just rely on AI even with these

tools. The last thing is rewarding attention. So how do we do better? We know that whenever an ending feels really good that people they feel like they got rewarded more. It's just like that buildup. This is a tool that helps you customize music specifically for the video that you're making. And uh this is all through AI. We couldn't do this before AI. You had to shift through

a lot of content. Now the gong indicates that uh this is the end, but I want to show you one thing for the end. So skipping through the slide, the one final use case, and that's the one I really recommend you to do even if you're not in the content industry. This has been the one I teased two years ago on WebET. Here I talked about this.

It's AI dubbing. AI dubbing. If you don't do this yet, I'm going to say it in the way I would say it to my team. You're absolutely stupid. Why? Because in the content creator space, we often get 400% more views once we go AI dubbing. And that's the localization of thumbnails, titles, and even our own voice. Now, I'd recap, but we're out of time. And recapping is

actually bad for engagement. But I want to thank you all and uh hopefully I gave you an insight in a little bit of my industry. Thank you very much.