A debate is brewing over the risks and rewards of serving ads to AI agents

Ads designed to influence AI agents rather than human consumers are emerging as a new marketing channel. They're also opening up a new brand safety battleground.

The big outstanding question: Once the AI platforms get wise to brands advertising to their crawlers, will they play ball or try to stamp out the practice?

Here's an example of how these ads function: Adtech startup Mobian recently worked with the publisher Time to serve "Agent Ads" for brands including Ally Bank and the Project Management Institute.

Previously, Time had created stripped-down, text-only versions of its webpages that AI agents could read more efficiently. Time then directed AI crawlers to read those pages — which human readers don't see — rather than the normal ones with design elements like images. Now, Time is placing ads on those pages. The ads render as blocks of FAQ-like text about the brand, optimized for the AI bots to ingest.

These ads could help companies leverage the "authority" of publishers like Time to impact how their brands show up in chatbots like ChatGPT and Gemini.

Ally Bank CMO Andrea Brimmer told me she was quick to experiment with advertising to AI agents because "the signals are impossible to ignore." Consumers visiting Ally's website from AI platforms were 3.5 times more likely to open an account than those from traditional search, and its AI-driven traffic is up 9 times year over year, she said.

"When the way people discover brands is shifting fundamentally that fast, I'd rather be in the room helping shape what comes next than standing outside it," Brimmer said.

The agent ads experiment has sparked a fierce industry debate on X and in my WhatsApp chats.

Among the talking points: Is this the start of a legitimate new advertising channel — or the next black-hat arms race? Could bad actors spin up their own agent ad networks to push scams or manipulate AI recommendations against competitors? And if they do, will the industry move quickly enough to establish standards before AI companies simply start blocking them all as spam?

There's a reputational risk if brands surreptitiously put their thumb on the scale. The industry has been here before. Bell Pottinger became a cautionary tale in the early 2010s after an investigation found the PR firm had covertly edited clients' Wikipedia pages to present them in a more favorable light.

"When the backlash comes, and it will come fast and furious, do you want to be the poster child for that backlash?" Bernard Urban, of the consulting firm BSCI, told me.

PMI CMO Menaka Gopinath said that, like much of online media, AI raises questions about transparency, accuracy, and the sources of information. She added that she appreciated how Time and Mobian approached the pilot, ensuring that sponsored content was clearly labeled.

"Organizations, publishers, and technology providers all have a role to play in building that trust," Gopinath said.

The most popular AI platforms have variously said in their policies and public statements that ads don't influence their answers — though this language appears to refer to the ads they run within their own chatbots rather than ads on websites targeted at their crawlers.

Mobian CEO Jonah Goodhart said that, broadly speaking, it's clear AI models are influenced by advertising.

"Models are trained on what people write, publish, and discuss, and advertising shapes both culture and the broader public record," said Goodhart, an adtech veteran who sold his previous startup, Moat, to Oracle.

With its ads, Mobian provides what Goodhart describes as "clearly labeled, brand-verified facts, backed by sources." How much weight the agent gives that information depends on the question, context, and the credibility of those sources, he said.

Ally's Brimmer told me that every new technology brings new responsibility, something she doesn't take lightly. She said that each ad aimed at LLM bots goes through the same legal, brand, and compliance reviews as everything it puts in front of people.

"I think there is a far greater brand safety risk in not paying attention to what LLMs are telling people every day about your company," Brimmer said.