Trust in an AI Video World

Thank you for the warm welcome, lovely to see you all here, thank you for joining this session in this amazing place. As Monique said, AI, video and trust, the question is can we have all three together? Or do we have to compromise on two of them? I'm going to try and do some research, some thoughts and some insights that hopefully will answer that question by the end of the session. There is a principle in neuroscience that says that the human brain makes a trust decision about a human face in 100 milliseconds. That's literally the blink of an eye. And that happens before language, that happens before conscious thought, and that happens before your carefully crafted marketing message has a chance to even land. Now, what if the human, or the face that your brain is judging isn't actually human? In January 2024, in Hong Kong, an employee of a multinational global organisation working in the finance team joined a video call with his CFO and with various other colleagues from the finance team. And as they went through that call, as part of the due diligence they were going through, he transferred 25 million US dollars into a different bank account. You can see where this is going, right? He didn't know, and his brain didn't know, that every single person on that video call was a deepfake AI. He had no idea. 25 million US dollars gone. Now, his brain, your brain, your audience's brain, all run exactly the same software. So how can organisations build genuine trust in an AI video world? That's what we're going to explore here in this talk today. In 1970, the Japanese roboticist Masahiro Mori noticed something interesting, also slightly strange. As robots were developing and becoming more human-like, more lifelike, people liked them more. Little robot arm, you know, doing work, it's quite cute, had a face and a head to it, even better. And then something strange happened. As these robots became more and more lifelike, but not quite human-like, the trust graph dropped significantly, very suddenly. People didn't just feel neutral about it at that point, they felt uneasy, they felt really disturbed by that kind of almost lifelike, but not really human-like resemblance. So the brain was firing a warning signal, and he called it the uncanny valley. And the reason it's a valley, rather than just a cliff, is that once you go beyond the valley, once you achieve almost perfect human-likeness, in robots, the trust actually returns. And the objection to what you're seeing, what you're experiencing, disappears. Now Mori was talking about robots, but 55 years later, that graph and what he described, sort of uncomfortably, closely describes how we feel about AI video today. In particular, the synthetic humans in AI video as well. And the issue is that most B2B marketers, I'm sure there's none of you here, but most other B2B marketers are building their AI video strategies exactly at the bottom of that valley, where the audience feels most uncomfortable. And that's a problem. Now, the brain has dedicated hardware to judge human faces and human voices, and it's one of the most ancient and finely tuned processing methods, that we possess. Millions of years of evolution have made it incredibly sensitive, because missing a warning signal, guessing it wrong, trusting the wrong person or the wrong situation, could be a survival problem. Now, what neuroscience tells us is that when something is almost, but not quite human-like, two different systems in our brain fire at the same time. There's the familiarity system, that says, I recognise this, it's a human form, or human face. And then there's the threat detection system as well. That says, something is not quite right here, just be careful, I don't quite trust them. And when they both fire at the same time, it creates this cognitive dissonance, this feeling of unease, that is beyond any sort of rational scepticism, and it's more sort of gut-level discomfort. You can't really describe why it is, but it happens because the brain is firing at the same time. The viewer does not have to know or think, this is AI that I'm watching. It happens before that, before even the recognition or the realisation that this is something fake. So the brain preempts all of that and goes way beyond that. Now I want to show you something that I found remarkable, genuinely remarkable. This is the uncanny valley that we looked at before. And then next to it is the Gartner Hype Cycle. Now, many of you will know the Gartner Hype Cycle, I'm sure. It describes the market's emotional journey with new technology, new emerging technology. We start with the innovation trigger, then we go into the peak of inflated expectations, then we drop down into the trough of disillusionment, and then afterwards we climb out slowly into ultimately the plateau of productivity. Now, look at both graphs together. They were developed independently, decades apart, three decades apart, 30 years apart, by completely different fields. One describes our physiological response to synthetic humans, and the other one describes the psychological journey that a market goes on with new technology. But they are both almost exactly the same shape. And that, I would argue, is not a coincidence. Both curves are describing the same human phenomenon, what we anticipate and what we are actually willing to accept. The peak, the crash, and then the recovery again. But they were told by two very different disciplines that aren't related. And all of this, and I've done a lot of research, the Gartner Hype Cycle was not based on Mori's uncanny valley. Now, where's AI video today? The numbers look impressive, until they don't. Gartner's 2025 analysis showed that AI video and generative AI was at the peak of inflated expectations, now sits in the trough of disillusionment. In 2024, on average, enterprise organisations invested 1.9 million US dollars each in Gen AI. But, less than 30% of CEOs were happy with the return on investment that they got from that. Meanwhile, we have 41% of people, video professionals, saying, yes, they use AI in their video workflows, in their video processes. And the video, AI video market is scheduled to grow, or predicted to grow, in the next six years, fivefold. From half a billion US dollars to 2.5 billion US dollars. But this is where the uncanny valley has a potential much bigger impact than the Gartner Hype Cycle. And I think it's probably one that perhaps most marketers have missed. We may not be moving towards a trough of disillusionment. We may be moving towards a trough of trust. Or let's call it the trough of mistrust. And that's a much harder problem to climb out of, and a much harder thing to overcome. Now, sadly, us humans are not quite as smart as we think we are. And I want to share something with you that I think should unsettle everybody here in this room. In the best possible way. A few months ago, I asked my LinkedIn audience, which is not massive, but enough to ask people for their opinion, what they thought about synthetic human faces, AI avatars, in video. And over 80% of the respondents said that it was creepy, it was just a bit weird, it's just not very good. And I also gave them the opportunity to tell me, would you be able to detect the difference between a human face and a synthetic face? Zero. Nobody said that they wouldn't be able to tell the difference. Everybody was positive that when faced with it, they would be fine, they would be able to tell the difference. A 2024 meta-analysis of 56 different studies, with over 86,000 participants, all investigated whether humans, people, would be able to tell the difference, and would be able to tell AI video, or deepfakes. The results of all those studies, on average, was that it was barely better than chance. The detection accuracy was 55%, about the same as flipping a coin. Yet, I think, probably everybody sitting in this room, you probably would have put yourself on the kind of the way past the 55%. Probably you can detect it, and then when you see it in your feed, you can probably tell. But the reality of this study shows that it's much harder than we might think. So, there's a different part of those studies that came out that I think is even more interesting. And that showed that several studies show that actually the response of humans to highly-optimised synthetic images was actually positive in many cases. So the reality of a human being was not quite as trustworthy as the optimised human synthetic face of an AI. And that shows the path towards something that I think that's much more meaningful and much more interesting. So we have a paradox in a way. We have the brain that actually is primed to detect and has that sort of cognitive dissonance when something isn't quite right. Detect that, you know, authenticity. Triggering the kind of uncanny valley and triggering that unease. But at the same time, we also have a brain that occasionally thinks that synthetically optimised faces are more trustworthy than the original human face. Now, in my mind, that's not a contradiction, but it is a warning signal, a warning sign, of the instability of trust in a synthetic media environment. And there's another force entering the picture. And I think that will define the next 18 months of B2B AI video. And researchers have a name for it. It's called the imposter bias. And the imposter bias describes a growing tendency of people to doubt the authenticity of digital content, including content that is entirely genuine. And you may have experienced this as well before, where you look at your video feed on whichever social media platform you're on, and you think, well, yeah, I think that's fake. I'm not sure. It could be real. Who knows? That's the imposter bias. You give the doubt to, actually, this could be fake. And as AI-generated media becomes more prevalent, people are starting to apply that general sense of scepticism to everything that they see, all digital content that they see. And that includes synthetic content and accurate, real human content as well. So we've established that they largely cannot tell the difference. People can't tell the difference simply because they can't rule out that it's fake. So for B2B marketers investing in real human content, making the decision to, no, I'm not going to go with AI. We're going to invest in this real human content, that should matter enormously because the imposter bias is eroding trust in that work as well. Now, I've got three pictures here of more or less liked people. We've got the US president, we've got the Danish prime minister, and we've got the very popular British prime minister. Just by a raise of hand, please participate. Don't hold back, don't be neutral, don't sit on the fence, but just raise a hand if you think, in this case, let's start with Donald Trump. That picture was taken from a real video or from an AI video? So who thinks that was from a real video? Hands up. Okay, who here thinks that was from an AI video? A few less. There's still a few. Okay. You know him so well, don't you? You see him everywhere, so you can probably tell. What about the British prime minister? Who here thinks that was taken from a real video? Hands up. And who thinks that was taken from a fake video? Okay, a few more. And then, Mette Frederiksen, Danish prime minister, who here thinks that was from a real video? And who here thinks that was from a fake video? Wow, okay. So arguably, the person that you should probably know best had the most votes for this is a fake video. And I can kind of tell what you think then. And I can tell you that all three, I know it's still from, you know, you haven't actually seen the videos, but all of these were taken from real videos. But your reaction shows the imposter bias. You can't quite be sure. So we start looking at different things, we start interpreting the kind of, you know, the blurriness of the background of Mette Frederiksen, or, you know, we look at, is that Trump how we normally see him? Yes, that's probably real. But the imposter bias is there, we can't be entirely sure, and we apply that to everything that we see these days. And that's incredibly important to us as people who build brands. So there's one final thing that I want to add to this as well, because it's really important. Just while I'm talking, have a look at the cat's thumb. Just in case you're wondering if it's real or not. Research published in Communication Psychology in 2025 showed that even when people were told that something is fake, even when the, like here, there was warning signs everywhere, their brain wasn't able to prevent an emotional reaction, an emotional imprint on the person watching it, even though they knew exactly that it wasn't right, and it wasn't true, and it was fake and AI. So the label saying that this is fake, it simply wasn't enough to counteract that. Now, I want to show you this video to explain a little bit about what I mean. We all had some kind of emotional reaction to this. We either felt sorry for the dog being blown up, or we felt sympathy with the cat for being so clever and laughing its head off. We all had a bit of a chuckle. We all had an emotional reaction. And that's fine when it's AI dogs and AI cats. But what if it's about the Iran war? What if it's about the Ukraine war? What if it's about your kids? Or about the company that you work for and the brand that you've helped build for the last 10 years? What if it's about you? The brain simply cannot make that distinction. And so when video, even when it's fake, lands emotionally, and the brain can't tell the difference, or can't make that distinction, and it processes that emotional content, that when it lands emotionally, regardless of whether it's fake, whether you know it's fake or not. And that should matter enormously to us. But it's also the most simultaneously exciting and terrifying and ethically significant thing that I'll say today. So, let's add one more layer, because we haven't got enough layers yet. Because the B2B video audience is changing. It's not static. Right now, Gen Z makes up approximately 18% of the target market of the workforce. But by 2034, Gen Z and millennials combined will make up 80% of the workforce. So that means that buyers, people who influence your deals, procurement managers will come from that generation. At least 80% of them will. They will increasingly grow with... There will be people who increasingly grew up with AI. Now, there's a common assumption in B2B circles that, you know, younger generations are more comfortable with AI content. But I want to challenge that here a little bit today. Because research consistently shows that Gen Z is the generation that is most finely attuned to authenticity and most allergic to its absence. So this generation is coming into the marketplace and it has, by a factor of five to one, it just does not trust companies to act in the best interest of society. And 58% of them don't believe that brands who are making claims can necessarily prove them. So they've grown up... Gen Z has grown up with influencer culture. It's the first generation who's really grown up with pervasive influencer culture. So they have developed an extraordinarily sensitive bullshit radar. Especially when they're being sold to, rather than being spoken to. And this generation is entering the B2B market primed to accept... Sorry, not primed to accept synthetic content, but to question it. And they won't disregard, they won't ignore AI content on principle, but they will call it out. Anything that seems manufactured, they will call out, and they will call bullshit, and they will say, that's not right, and that's not real. And that's going to be 80% of the workforce by 2034. So for Gen Z and millennials, the uncanny valley, is not about robots, like it was for Masahiro Mori. For them, it's about trust in every single brand represented in this room here today. So for us, that should matter enormously, and it's something that we should care about, to think about, okay, who's our target audience going to be, and what are we going to do with the content targeting that target market? So where does all this leave us practically? I want to give you a sort of quick, easy, mental model, to take away with you from here today. When you think about your AI video decisions, I'd like you to think about them as sitting in three different zones. Zone one is before the valley. There is no pretence here of being human. We're talking fully synthetic, but nothing human. So we're talking about data visualisation, motion graphics, animated explainers, AI generated B-roll. These are things where there's simply no human face involved. So the brain processes this as a tool. I don't really care. There's no valley being fired off here, no kind of cognitive dissonance. Safe. And if you're not operating in this space right now, adopt it. Immediately, heavily, completely safe zone. Absolutely fine here. Then we've got zone two. Zone two is the danger zone. Proceed at your own risk, and proceed with extreme caution in this. This is where we're talking about realistic AI avatars, synthetic humans, synthetic moderators and presenters. We saw the judgment call the brain makes 100 milliseconds right at the beginning. That happens straight away when they see something like this. There's the imposter bias, so people are already confused as to what's real and what isn't. And then that emotional response. The thing is, if you adopt this space, if you go into this space, no matter how many warning signs you put up, like with the cats and dogs, the brain will have an emotional reaction to it. And because it happens in 100 milliseconds, you have no way of controlling that. And you don't know. It's super unstable. You don't know whether it's going to be positive, neutral or negative. Yes, some people might like it. But who is going to lose trust in your brand because they don't like it? So this is an area that I just caution you again. Proceed with extreme caution. And it may be right, it may not be right, but just be aware of how the human brain interacts with it. And then there's zone three. This is beyond the valley. This is a safe zone. Again, you see, in safe territory. This is things like AI-assisted script writing, AI-powered repurposing, text to subtitle, AI processing, things like that. Again, you're not putting a human face on. You're using it as tools. But you are creating human content that is supported by AI. Again, a safe zone. You're not replacing this. The human face on the screen is real. The voice that you're hearing is real. This is another safe zone. So you're fine. So be aware of zone number two. So the shift I'm asking you to make is not a complicated one. Before your next AI decision, before you deploy or before you even brief in a realistic AI avatar, I'd like you to ask a different question. Don't ask, can AI make this content? But instead ask, what does a human brain actually need from this content? And is AI the right instrument to do it? And sometimes the answer will be, yes, that's fine. We can do it with AI. And sometimes the honest answer will be, no, actually, we need a real human face. That actually means that we need to be human in this content. And it's for you to make that decision, being mindful of what the trust response will be. So I want to leave you where Mori began. He wasn't trying to stop robotics. He was a roboticist. He wasn't anti-technology either. But he was trying to make engineers who were developing these robots smarter about the human response. He was saying, the machine is only as good as the human experience it creates. Let me say that again. The machine is only as good as the human experience it creates. And that is our job as B2B video marketers today. Not to resist AI, not to embrace it uncritically either, but to be the people in our organisations who ask the right question. Who understand what the brain on the other side of the screen actually needs. And we need to apply the question, what does the human person, the individual need from this, before we ask, what can the technology do for us? So the uncanny valley is not a wall. It's not a cliff either. It's a valley, but it's also a map. And it helps us to navigate. And it helps us show the way where we're navigating to next. And it also helps us to navigate or circumnavigate perhaps certain areas that we maybe need to be careful of. So I wish you all the best on that journey. And if you like, I'd love to hear where you are on that journey and where you are and how you are navigating that map. In the meantime, it's over to you to navigate that space. And I wish you all the best. Thank you.