This agency’s AI detector tool could help brands monitor for deepfakes

Volvo, Bombay Sapphire, Chanel—each of these brands have been emulated, without permission, in AI-generated content made to look like real advertisements. The spoofs tend to be harmless, their creators readily admitting they have nothing to do with the actual brands. A thin line exists between a fun experiment and an act of trolling that could trick consumers and cost marketers severely. “You know it’s going to get worse,” said Chris Graves, chief creative officer at Team One, a Publicis Groupe-owned ad agency. This is the context in which Team One has launched Faikcheck, a web tool for quickly detecting whether content is real or a product of AI. A user uploads a piece of media, Faikcheck scans its details and outputs the verdict, including a summary of its conclusions and a confidence score up to 100%. The tool can only scan images for now, as Team One considers how to do the same for video. While Faikcheck is meant to help all audiences, not just marketers, the ad industry is in growing need of solutions that identify AI-generated content. Advancements in image and video technology are eroding the creative barriers that have previously protected brands from having their intellectual property reconstructed in fraudulent yet believable ways. Video generation platform Showrunner has developed TV pilots based on Disney IP, while marketers themselves are hiring companies like Runway to build models based on their copyrighted material. See all of Ad Age’s 40 Under 40 honorees here Brands are behind the curve, Graves said, in part because the emerging AI detection tools have largely focused on text, leaving out the kinds of visual media that most commonly deceive unwitting viewers. Typically known as deepfakes, this misleading content has advanced in lockstep with the rising quality of AI video generators. For example, Meta this month launched a stirring video platform that, among other things, can generate a video of a real person doing virtually anything based on a single image of their face. What brands can do To be sure, Faikcheck is far from the first AI detection tool capable of processing images. But its development by an ad agency is relatively novel, and reflects a frustration felt by Graves that AI providers aren’t doing enough to help curb the flow of AI-generated misinformation. Instead, marketers have had to pick up the slack, such as consumer health company Haleon, which has adopted digital watermarking technology to authenticate its brands’ ads. Another worry facing marketers, Graves said, is simply appearing next to AI-generated misinformation. Indeed, in a 2024 consumer survey conducted by IPG Mediabrands, only 36% of respondents found it appropriate for brands to appear next to such content. This concern increases the need for brands to identify material that may be fraudulent so that they can act accordingly, such as by applying misinformation reporting to their ad campaigns or leaving the platform in question altogether. At the same time, marketers have to learn to let go of some control, Graves said. The fear is that one’s IP can now be exploited in both detrimental and believable ways. But Graves stressed that, in most cases seen so far, the creator is not profiting off of this fake content, and thus little retaliatory action has been required. “Brands used to tell consumers how to interpret their IP, but now people on social media are flipping that on its head,” Graves said. Tools like Faikcheck will be important as the flood of AI-generated material surges on the internet, but these solutions, too, must be taken with a grain of salt. Faikcheck has shown a tendency to conflate heavy editing with evidence that the image was AI-generated, Graves said. Faikcheck itself is built on top of ChatGPT, which perhaps helps explain this imperfection. “AI would be the one technology to know how to do this because it knows itself,” Graves said.

This agency’s AI detector tool could help brands monitor for deepfakes

Featured Stories