Gen AI Can Tailor Ads to Our Personalities—and They’re Pretty Persuasive

Now, imagine that every ad you see online speaks to you like this. It’s almost as if the advertisers know you, your personality, and the things that matter to you.

Welcome to personalized advertising in the age of generative artificial intelligence (gen AI). Advances in gen AI and our expanding digital footprints are not only making highly personalized ads possible, they’re making them more likely, says Jacob Teeny, an assistant professor of marketing at the Kellogg School of Management.

“This specific influence tactic, personalized persuasion, is now going to be more scalable and widely deployed than ever before,” Teeny says.

By using free applications like ChatGPT—along with personal information about our browsing behavior, social-media posts, smartphone use, and more—advertisers can create highly personalized and therefore influential marketing appeals.

To learn just how effective these customized advertisements can be, Teeny teamed up with a group of researchers including Sandra Matz, Heinrich Peters, and Moran Cerf of Columbia Business School; and Sumer Vaid and Gabriella Harari of Stanford University.

The researchers set up a robust series of studies that gauged how consumers reacted to personalized pitches written by ChatGPT. Overall, they found that this gen AI was able to write messages tailored to people’s complex psychological profiles, and these messages were more persuasive than non-tailored or generic messages.

Personality traits influence ad preference

In an initial study, the researchers showed 120 participants pairs of messages advertising an iPhone and asked them which message they preferred. One message was designed by ChatGPT to appeal to someone who scored very high on a given personality trait, and a second message appealed to someone who scored very low on that same trait.

For instance, an excerpt from the ad for extroverts read: “If you’re the life of the party, always up for a good time, and enjoy being surrounded by people, then this is the phone for you!” Meanwhile, an excerpt from the ad for introverts read: “If you’re looking for a phone that won’t draw attention to itself, the iPhone is the perfect choice.”

Participants then rated which message was more persuasive and which made them want to get an iPhone. They then took a 30-question survey to identify their own personality traits.

The researchers found that for three of the four personality traits tested, the personalized ad designed by ChatGPT (as determined by whether participants scored high or low for a given trait on the personality survey) was more persuasive than the non-personalized ad.

Consumers don’t mind being targeted

But how would people respond if they knew the message had been personalized to them using gen AI?

For the next study, 471 participants completed the same tasks described above but with a wrinkle. One group of participants served as a baseline group, while another group was told that gen AI had created the ad for them based on their personality traits.

Again, the researchers found that the personalized ads were effective overall, replicating the original finding. But they also found that the participants knowing the ads had been generated by gen AI to appeal to them specifically didn’t change the ads’ effectiveness.

That’s notable because prior research has shown that people try to resist being swayed when they know someone is trying to persuade them, Teeny says.

“We’ve seen a bit of cultural shifting in terms of the acceptability of being persuaded. When we go online, we know we’re getting targeted ads. And because that’s just become the norm so much, I think people are more OK with it,” Teeny said.

Simulating how advertisers use AI

Finally, the researchers wanted to use what they had learned and put it into action—much in the same way that advertisers would use consumer data. So, they invited 297 of their earlier participants back for another study more than half a year later.

For each participant, they relied on their previous data to identify their most salient personality trait (out of extroversion, openness, and conscientiousness) and then used that to have gen AI create new personalized ads, this time promoting a Rome getaway or a pair of Nike sneakers.

Participants were shown both ChatGPT’s personalized and generic ads for each offering and rated the persuasiveness of the ads as well as how much they would be willing to pay for them.

This time, using messages that were personalized to each participant’s personality from data collected half a year ago, the researchers continued to find that the personalized ads from gen AI were more effective than the generic ads. This was particularly true for the trip to Rome relative to the sneakers.

This makes sense, Teeny says, since someone’s personality might be more at play when selecting experiences (which can speak to many of a person’s traits) than a consumer product (which may speak to fewer). After all, extroverts and introverts both wear shoes, but will likely have varying preferences for how they’d like to spend their free time.

“The more that there’s an opportunity to tailor the service or the product to a person’s predisposition, the more that generative AI or even a human author should be able to tailor a message to appeal, and thereby influence, your interests and behavior,” he says.

Don’t believe everything you read online

All of this sends a clear message to marketers: personalizing messages to consumers is greatly effective, and free AI tools are making it easier than ever.

For consumers, the flipside is that this new frontier of messaging requires everyone to be even more mindful of the content they read online, Teeny says.

“We’re going to be inundated with things that naturally feel like they appeal to us. So we might have to take a second step to really investigate the source or the veracity of the message—whether that’s in regard to a consumer product or in regard to a political news article,” he says.

Not taking that extra beat to stop and think might mean buying a product you really don’t need or voting for a politician who never endorsed the messages attributed to him or her, he says.

From a policy standpoint, the research suggests that social platforms may need to do more than simply flag AI-generated content to protect users from being persuaded by misinformation. “Meta is implementing a new policy where they have to say if an image was created or an ad was created with AI. I think that’s a great first step, but as we see in this paper, that doesn’t necessarily resolve the issue,” he says.

As gen AI continues to improve and evolve, it will create even more persuasive content, Teeny says. The study was done using ChatGPT-3, which is already outdated, and it’s likely that newer versions will get better at incorporating more facets of someone’s personality in a single ad.

“The effects are probably only likely to get stronger as time persists,” he says.