The Great AI Advertising Reset: DoubleVerify CEO Mark Zagorski on AI Slop and Fraud

It is great to be here and I'd say it's a privilege to have my first discussion with AdTech God. Thank you for making time for me today, Mr. God. Can I call you Mr. God? Or what should I call you?

It's a culmination of every little boy's dream in AdTech to have both you and Jeremy on the same call.

Absolutely.

We've got Mark Z with a Mike D.

Neural is a very exciting launch. But I call it a launch, but it's almost like it's an 18-year-old, right? It's been 18 years in the making, so it's kind of like a launch to your kid to college. Neural is the brain. It's the AI functionality that's behind a lot of what we do, whether it's measurement or optimization or our new agentic interfaces or our way of dealing with the challenges of AI slop. Nora is the brain. It is the core processors through which we deliver our solutions and those have been built since the day this company started. Before we started calling things AI and naming our AI cute names, we were building machine learning models and those machine learning models evolved and now they're being deployed through agents and now they're being deployed to attack and deal with agents. So, Nora is really the culmination of a lot of what we've done over the last 18 years.

There's so much noise around agentic advertising and agentic buying and selling and I think the definitions have become very loose, but the actual implementation is still very early. Most of the work is still being done in silos and I think DV's role has always been one in which we want to drive trust and transparency and break down those silos. So, we've leaned into different protocols, whether it's AdCP or with what's going on with the IAB. So, our first take on the agentic side is the answer is yes. We're building for all different agentic environments because we have to, because our customers will be playing all of them. But the second is we really think we want the agentic world to be as transparent and trustworthy as hopefully the adtech world has evolved to be, which is you need to be identifying agents. You need to be part of an ecosystem in which there are protocols. Those are things in which we are going to play an active role in and in the same way we verify sellers of media, context of media, and quality of media. We are now leaning into with products like Agent ID and other solutions that we've launched, identifying and qualifying agents. I think that will be the next opportunity for independent verification.

It's pretty stunning how rapidly the AI slop universe has become. I read a stat yesterday that said 70% of images in social media are AI produced now. But it's a massive amount of both contextual inventory, video inventory, and images that are being created by AI and advertisers are notably concerned. This is a huge issue. If you're a high-profile luxury advertiser, do you want to be around images of people who have an extra digit? Some AI slop is considered okay, some it's not. Most of it's MFA. We called this something else three years ago before AI was part of it. It was MFA, now it's AI slop. So it's an issue. We've built this tool called Slop Stopper to help advertisers avoid ads. Great name, we love it. I didn't come up with it, I have to give credit to my colleague Jack Smith who came up with that name. It keeps advertisers away from slop or low-quality AI-produced content in the open web. We've launched that across YouTube now, which gives them the ability on a pre-bid side to avoid low-quality AI-generated content in video. So it's a real challenge. We're building tools for it. We're using AI and Nora as the brain to identify that stuff. Using AI to fight AI. In the same way slop is a problem, AI is now fueling more and more fraud. This quarter alone in Q1, we saw a 140% increase in CTV fraud. Most of that now because an AI engine can generate 10,000 websites in a matter of minutes. You can create fraudulent CTV apps that you push into app stores. Before they get caught, they could be downloaded thousands of times running fraudulent CTV engagements. So AI is fueling growth in fraud, growth in slop, and making things sometimes easier for advertisers but mostly harder and making less transparency in the places where they used to buy.

Part of the challenge with any new solution is show me what it does. Give me the value. Show me how it works. Deliver me results. The great thing about what Nora brings to bear is it's actually just fueling results that we've been providing to advertisers for years. We can show the results. We can show you the fraud that we're removing by using AI. We can show you the slop that we're able to pull out. We can deliver results on optimization that are clear and show real ROI. So the value prop behind Nora is an enhanced version of the value prop we've been providing to our advertisers since day one, which is helping them find better quality media, eliminating waste, and doing so at a more efficient cost, but doing that faster, more accurately, more seamlessly, and now in a way that's more natural to them, using natural language interfaces for advertisers to be able to put in their brand suitability or brand safety cues. The ability for them to work in the agentic universe with safety and confidence by leaning on our solutions. So it's really, don't believe what we say, believe how this works, and it's working better for you, Mr. Advertiser, as you employ these tools.

I kind of imagined it before, but I think speed - insight to action faster. The ability for us to close the loop, understand something, and give advertisers the ability to act on it quicker. More confidence, the ability to know that when I am advertising, whether it's across social or CTV or the open web, that I can do so with the confidence that there's someone who has my back, looking for AI slop, looking for these environments that are challenging or trying to protect me from fraud. And then finally, more control. As we move towards an agentic universe where advertisers will control that agent and have the ability for that agent to really follow its own directions and push out specific granular preferences on things that it does or does not want to buy. The ability for us to work with that ecosystem with them. So really it's speed, confidence, control. Neural powers all of those things. It just makes DV better and it makes our customer engagements better.

Thank you.

Good stuff, guys. Thank you.