Maybe advertisers should rethink their involvement in anything that monetizes the attention of kids and teens online – especially as it pertains to new AI chatbots and software.
That’s according to former FTC Commissioner Alvaro Bedoya, who shared some of his qualms with digital advertising’s more controversial targeting tactics at Programmatic IO in New York City on Monday.
Bedoya called out social media and generative AI platforms for striving to keep young people engaged without basic safety guardrails in place. He also criticized digital advertising’s invasive tracking of highly sensitive data for users of all ages. He even cast shade on the concept of consent-based targeted advertising, arguing that the ad industry’s data-gathering practices should be completely reevaluated.
But Bedoya also had practical advice for how advertisers can avoid drawing the ire of tech regulators at the state and federal level, particularly as AI products inevitably turn to ad-supported models.
Keeping AI in check
Bedoya has some credibility to speak on these topics as a longtime privacy advocate in both the public and private sectors.
He has served as a senior adviser to the American Economic Liberties project since he and his fellow Democrat-appointed commissioner Rebecca Slaughter were removed from their FTC positions by President Trump in March. Their firing violated nearly a century of legal precedent regarding the FTC’s institutions – precedent that will be argued before the Supreme Court later this year.
Prior to his three years as an FTC commissioner, Bedoya founded Georgetown Law’s Center on Privacy and Technology and was the first chief counsel for the Senate Judiciary Subcommittee on Privacy. He’s had a hand in shaping federal tech policy and seen reams of research and expert testimony on the potential perils of consumer-facing technology and media.
At Prog IO, Bedoya saved his harshest criticism for the lax safety standards of “parasocial” generative AI chatbots like OpenAI’s ChatGPT and Google Gemini, particularly when it comes to how they’re marketed to and used by minors.
He called out instances of chatbots encouraging teenagers to commit suicide. He also pointed to research showing that people of all ages believe they can build intimate connections with chatbots and that some children find chatbots to be more trustworthy than other people.
Many chatbot operators allow services to represent themselves as psychologists or therapists or even release their own products with that branding, Bedoya said. However, humans who work those jobs require professional certifications, and kids and teens in particular can be easily fooled into thinking a commercial chatbot product is akin to a professional.
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
“Children, teenagers and adults are spending hours upon hours developing what they see as meaningful relationships with for-profit algorithms,” he said. “That is ripe for abuse.”
Potential harms could be mitigated by basic protections, Bedoya said, such as the chatbot sharing a persistent banner directing the user to a suicide prevention hotline if they’ve expressed suicidal ideation at any point in their interactions. LLM and AI companies could also show a persistent banner reminding younger users that “this is not an actual human being talking to you,” he said.
Bedoya dismissed the idea that reining in chatbots’ toxic incentives through regulation could stifle technological innovation. He added that, while some use cases for generative AI show promise, “innovation is not an unalloyed good.”
And even if tech regulation slows at the federal level under President Trump’s administration, Bedoya said he’s heard from the offices of multiple state attorneys general who are interested in continuing the FTC’s work on these matters.
With that in mind, Bedoya offered the following advice to the advertisers in the room: “Be very cautious about opportunities to advertise for teenagers and minors on these platforms,” he said, alluding to those AI chatbot and search engine companies. “I would not want to get mixed up in this technology when, in my view, the creators have not taken the most minimal steps to protect the young people who use it.”
Parasocial media
AI chatbots, like social media, crave user engagement and attention. That’s their North Star.
So, Bedoya said, advertisers have an important role to play in steering social media and generative AI companies away from harmful practices.
Americans have grown incredibly distrustful of social media and Big Tech giants and are outraged at how these companies treat young people, he said. He shared anecdotes of speakers coming before the FTC to testify about completely unrelated regulatory matters, and then pulling him aside and imploring him to “do something about social media” because of “the pain this is causing in my kids’ lives.”
The main thing advertisers can do to foster consumer trust in online experiences, he said, is to “think long and hard” about how they spend money with companies that strive to “keep teenagers online for longer and longer and longer periods of time.”
When it comes to generative AI experiences, Bedoya said, marketers should avoid repeating social media’s mistakes in maximizing engagement at all costs.
He said he understands, from a marketer’s perspective, the impulse to reach young audiences early and to establish brand loyalty before people actually have discretionary income to spend.
But, he added, given the bulk of research that shows extended social media use is detrimental to young peoples’ mental health, not to mention the “completely wild studies about AI chatbots,” it’s worth rethinking anything to do with targeting kids on the internet.
“I don’t think, where we are in 2025, that it’s a good idea,” he said.
Quitting consent
Advertisers shouldn’t just be wary of how they target young people.
Bedoya also pointed to worrying data collection practices for adults, such as mobile location tracking. He said the FTC examined cases “where there would be extraordinarily detailed information produced about people based on precise geolocation.”
He cited examples as extreme as tracking whether a user visits a specific endocrinologist’s office, then tracking whether they visited a pharmacy immediately afterward. “This is painfully sensitive information that these companies were producing and selling to the highest bidder,” he said.
He added that users often “had no way to express any kind of consent,” nor could they remove their data or profiles.
Indeed, putting the onus on consumers for opting into data sharing may also have to be reexamined in the long term, Bedoya said.
“The consent model itself is problematic,” he said. “There’s just too many yeses and noes people have to answer.”
