Mere months out from the removal of third-party cookies in Chrome, publishers might be finding themselves in a sticky situation. UK publisher video revenues fell by almost 18 percent last year, according to the Association of Online Publishers (AOP) and Deloitte, leaving digital publishers seeking richer monetisation methods for their video content.
One such model is contextual targeting, enabling advertisers to place ads that align with video content. And investors are picking up on the direction of travel; Qortex, an AI-driven contextual advertising company, raised $10 million in December, signalling venture capital interest in cookieless targeting alternatives.
At the same time, having AI as part of your pitch makes for an easier sell in the current climate, as Qortex Co-Founder and CEO Zack Rosenberg explains: “I don’t know that you can fundraise only with an advertising story. Without the AI story hitting at the right time, I don’t think there would have been a cheque.”
But Qortex only added the .ai to its domain last year. The company started life as FANTOM in 2017, a sports media business that acquired live rights to small, independent leagues, such as junior college football, Ultimate Frisbee, and even Major League Eating. But despite generating over 600,000 viewers per day on social media, the lack of monetisation options meant FANTOM was making just $11 per day in revenue.
As a result, Rosenberg developed an “On-Stream” video advertising product, launching as CatapultX in 2019. The next step was to build an AI system to recognise video content, rebranding as Qortex in 2023. “The journey of On-Stream was to solve my problem as a publisher; the journey of AI was to develop a system that understands what’s happening in the content,” says Rosenberg.
Scratching the surface
Qortex gained early credibility with its on-stream technology, whereby ads are overlaid onto a video, rather than interrupting the content. And sports proved a strong testing ground given the lack of natural breaks in games (and indeed competitive eating). The technology was soon rolled out to other forms of content, across web, CTV, and games such as Minecraft.
Adding the contextual component allows advertisers to target specific video content, for instance serving Uber Eats ads on top of a cooking scene. But Qortex’s Intelligent Video Analytics (IVA) technology goes one step further, according to the company, using AI to perform deeper contextual analysis.
“A lot of contextual tools would identify an aeroplane in a video,” comments Rosenberg. “But is it taking people on vacation? Or did something tragic happen to them? You have no idea. We cut the video up into 20- to 30-second chunks, we call them ‘moments’, and within those moments we contextualise what’s happening. And that is a huge leg up from a brand safety standpoint.”
Yet despite the obvious benefits of showing ads related to the content someone is watching, video advertisers have been slow to adopt contextual targeting solutions, with concerns around performance and attribution. Rosenberg argues that this reluctance is misplaced.
“You never heard Google complain that contextuality didn’t work for them,” he remarks. “If I type into a search bar that I want to travel to Belgium, all of a sudden ads for hotels in Belgium come up. Contextual relevancy has never seemed to fail them. The problem is that we haven’t been able to equate that to video. Advertisers have been discussing audience targeting for so long; now we’re just starting to scratch the surface on the concept of moments.”
Artificial artificial intelligence
The ability to explain the actual content of a video is helping to shift that mindset, according to Rosenberg. He notes that if the internet stopped accepting videos today, it would take 17,000 years to watch them all. And while AI can help make sense of that mass of video content, it will also drive its proliferation, potentially making content harder to decipher.
“Deepfakes will be one of the biggest problems of our times,” says Rosenberg. “Presidential candidates are probably the easiest example of being able to identify deepfakes, because everything they do and say is documented and available publicly. But if you’re a congressional candidate, and there’s not that much video of you, you don’t have that comparison to know this isn’t real.”
Rosenberg adds that AI companies like Qortex can help identify nefarious content from a brand safety perspective; “If you have Joe Biden getting on screen and talking about how much he loves Nazis, our technology can identify that this is hate speech, and therefore you as an advertiser don’t want to be associated with it.”
But he also calls out the difference between being an AI company and certain trend-hopping claims. “I call it artificial artificial intelligence,” he says. “Everybody’s saying that they have AI, but they don’t. They licence something from AWS and they’re like, ‘See, we told you we’re an AI company.’ But I don’t think that qualifies you as being an AI company; it qualifies them as being an AI company. There’s a lot of that out there as well.”
