GWI’s Jason Mander: Most Marketers have Encountered Negative Impacts From Inaccurate AI Results
The cliché about AI tools is worth repeating: the output is only as good as the context fed into them. That gap between capability and reliability is what GWI’s new research puts a number on. And the findings are striking.
“The overwhelming majority of marketing professionals have suspicions about AI capabilities and over 80% have found their AI tool giving a bad or inaccurate audience take. But hardly anyone has the time or inclination to check and verify that,” Jason Mander, chief insight officer at GWI, told Beet.TV Editorial Director Lisa Granatstein. “There’s a 40 to 50 point percent difference between the numbers who are worried, the numbers who are spotting mistakes, and the numbers who are actually verifying the insights.”
The research, titled Artificial Confidence, frames this as an upstream problem. Individual marketers cannot be expected to audit every AI output. The data flowing into these tools needs to be reliable by default.
AI only knows what you tell it
GWI positions itself as the third pillar in a three-part framework for AI-powered decision making. The first pillar is AI itself, where capabilities across tools like ChatGPT and Gemini are converging. The second is proprietary intelligence, what a company knows about its own customers and market. The third is human intelligence, the kind of representative, large-scale behavioral data GWI provides.
“That first pillar, artificial intelligence, doesn’t really know anything. It only knows as much context as you give it,” Mander said. “You need that third pillar of human intelligence to really make the best possible decisions that reflect how people actually do things, not just the stereotypes or unrepresentative data you might find on the open web.”
Speed and verification can coexist
The instinct to slow down and verify AI outputs runs into a practical wall. Marketing professionals simply don’t have time to fact-check every insight. Mander’s argument is that verification doesn’t have to be a separate step added after the fact. It needs to be built into the inputs from the start.
“Speed and verification don’t have to be in conflict with each other. They can come together. Your workforce needs to be able to trust the inputs going into the tools,” Mander said. “Around 80% of the marketing professionals we spoke to have encountered material negative impacts from inaccurate AI outputs. Whether that’s time lost at the lighter end of things, but it goes all the way through to reputational damage, customer complaints, and poor performing campaigns.”
That finding points to an industry ready for a different conversation about AI reliability, one that goes beyond celebrating efficiency gains to asking harder questions about what’s powering the outputs.
“It’s an industry that is crying out for reliable data sources that it can trust to feed in and power the AI tools which are being so widely used,” Mander said.
You’re watching Beet.TV coverage of Advertising Week New York 2026. For more videos from this event, please visit this page.
The cliché about AI tools is worth repeating: the output is only as good as the context fed into them. That gap between capability and reliability is what GWI’s new research puts a number on. And the findings are striking.
“The overwhelming majority of marketing professionals have suspicions about AI capabilities and over 80% have found their AI tool giving a bad or inaccurate audience take. But hardly anyone has the time or inclination to check and verify that,” Jason Mander, chief insight officer at GWI, told Beet.TV Editorial Director Lisa Granatstein. “There’s a 40 to 50 point percent difference between the numbers who are worried, the numbers who are spotting mistakes, and the numbers who are actually verifying the insights.”
The research, titled Artificial Confidence, frames this as an upstream problem. Individual marketers cannot be expected to audit every AI output. The data flowing into these tools needs to be reliable by default.
AI only knows what you tell it
GWI positions itself as the third pillar in a three-part framework for AI-powered decision making. The first pillar is AI itself, where capabilities across tools like ChatGPT and Gemini are converging. The second is proprietary intelligence, what a company knows about its own customers and market. The third is human intelligence, the kind of representative, large-scale behavioral data GWI provides.
“That first pillar, artificial intelligence, doesn’t really know anything. It only knows as much context as you give it,” Mander said. “You need that third pillar of human intelligence to really make the best possible decisions that reflect how people actually do things, not just the stereotypes or unrepresentative data you might find on the open web.”
Speed and verification can coexist
The instinct to slow down and verify AI outputs runs into a practical wall. Marketing professionals simply don’t have time to fact-check every insight. Mander’s argument is that verification doesn’t have to be a separate step added after the fact. It needs to be built into the inputs from the start.
“Speed and verification don’t have to be in conflict with each other. They can come together. Your workforce needs to be able to trust the inputs going into the tools,” Mander said. “Around 80% of the marketing professionals we spoke to have encountered material negative impacts from inaccurate AI outputs. Whether that’s time lost at the lighter end of things, but it goes all the way through to reputational damage, customer complaints, and poor performing campaigns.”
That finding points to an industry ready for a different conversation about AI reliability, one that goes beyond celebrating efficiency gains to asking harder questions about what’s powering the outputs.
“It’s an industry that is crying out for reliable data sources that it can trust to feed in and power the AI tools which are being so widely used,” Mander said.
You’re watching Beet.TV coverage of Advertising Week New York 2026. For more videos from this event, please visit this page.
Series Videos
