Interpersonalization

Interpersonalization

Does the marketing dream of AI-generated ads for every person make sense?

By Brad Berens

Digital advertising uses all the data it has about you (which is a lot) to draw conclusions about the things you want and then place ads about those things in front of you. My friend (and media deep thinker) Jim Meskauskas describes the worst-case scenario this way: “advertising is helping people you don’t like sell things to people you don’t know to buy things people don’t need with money they don’t have.”

Since the first banner ad winked onto www.hotwired.com, the holy grail for digital marketers has been 1:1 advertising: putting the right ad in front of the right person at the right time.

With the arrival of generative AI, the size of that opportunity has grown. Now, the dream is to have AI personalize the words and imagery of an ad for an individual consumer (a terrible word) in the picoseconds between when a person has indicated interest in some topics and the moment when that ad appears next to the thing they’re interested in.

Let’s say a cluster of ad-tech companies collectively know that Terri is a 34-year-old Black woman who: 1. lives in Chicago; 2. is an avid outdoorswoman who skis in winter and hikes in summer; 3. has a job that pays her more than $175,000 per year; 4. is unmarried but in a relationship with Chloe, a 37-year-old Asian woman; 5. has a four-year-old female Australian shepherd named Bean; 6. shops at Sephora and Nordstrom; 7. buys groceries at Whole Foods and Mariano’s, and 8. has lingered over ads for new SUVs for half a second longer than other ads. The companies probably don’t know Terri’s name or other Personally Identifiable Information (PII), but they have an anonymized version.

In a real-time, AI-driven auction, different SUV manufacturers will duke it out for the privilege of putting an ad in front of Terri. Subaru wins the auction. (IYKYK.)

At that moment, the digital marketing dream is for AI to instantly create an ad that features an upscale Black and Asian lesbian couple driving a Subaru SUV towards the slopes on a sunny day with skis on top and the dog’s head poking out of the window, tongue lolling in the wind. Terri sees herself in the ad, realizes she needs a new car, and puts Subaru into her consideration set. (It’s a lot more complicated, but you get the idea.)

Is this really how our minds work?

I don’t think so. I’m skeptical about how effective this sort of real-time personalization will be at persuading folks to buy things because—unless you’re Henry David Thoreau or Ted Kaczynski—identity is a social construct. Hyper-personalization is an example of over-focusing.

We calibrate ourselves in comparison to people we know.

For example, years ago I needed to buy new blue jeans. I had mostly worn Levi’s (except for a painful childhood chapter of Husky-sized Toughskins from Sears), but as I was shopping I remembered that my crazy friend Joey wore Lucky Jeans. At that time, Joey was a peacock, a flamboyant, fast-talking sales guy who charmed everyone around him and was always put together. (He has since become a family man, memoirist, and podcaster with whom I occasionally argue about how bad Jonathan Haidt’s books are.)

At the store, I thought, “I wonder how these jeans would look on me?” When I tried them on, I saw that the inside flap of the front zipper had a little vertical sign that read, “Lucky You”—a sign clearly intended to be read by another person in a specific posture. I laughed out loud, tried on the pants, liked the way they made my butt look, and have worn them since.

No advertisement, even one personalized to me, would have accomplished this.

Goodreads for the other parts of our lives

Instead of personalization, what AI could do to increase the quality of both advertising efficiency and our lives (imagine that) is by leaning into interpersonalization.

This means surfacing the experiences and opinions of the middle-tier of the people you know. Imagine a target with a bullseye and two bigger rings (the Target logo is an example).

The bullseye represents the people you already consult. I talk with Ben and Peter about books. I talk with Juliet and Pam about restaurants. I trade texts with Raman about apex science fiction nerdery. I don’t need an AI or an algorithm to remind me to ask them.

The outermost ring represents the people you know who have no taste, whose opinions you don’t respect, or whose lives are so different than yours that it’s amazing you can even have a conversation.

What interpersonalization can do is make thinkable all the people in the middle ring of the target: the folks whose opinions you do respect, but it doesn’t occur to you to ask them.

Goodreads does this for books: what are the friends I don’t often talk with reading these days? With AI-driven interpersonalization, this middle-tier word of mouth could expand to most other economic decisions.

Social media influencers do a version of this, but it’s parasocial rather than social: we don’t really know these people. Groupon, Living Social, and Gilt Groupe did discount-driven versions, but they were anonymous. Some airlines do a version with that irksome, “2,312 other travelers bought the insurance today” pressure tactic.

Since the first banner ad winked onto www.hotwired.com, the holy grail for digital marketers has been 1:1 advertising: putting the right ad in front of the right person at the right time.  With the arrival of generative AI, the size of that opportunity has grown. Now, the dream is to have AI personalize the words and imagery of an ad for an individual consumer (a terrible word) in the picoseconds between when a person has indicated interest in some topics and the moment when that ad appears next to the thing they’re interested in.

Interpersonalization would do it with people you know. It would also be more persuasive and less creepy than the “why is this ad following me all over the internet?” phenomenon.

Interpersonalization is no small technical challenge because it requires interoperability across different AIs, ecommerce sites, social media networks, and more. The privacy challenges alone are formidable. (Plus, we can’t even get Netflix to suggest something that two people will enjoy watching, so interpersonalization will take a while.)

Amazon could do this within its ecosystem, which is scary since Amazon already gets so much purchase-related attention, not to mention purchases. In 2017, when Verizon decided to shut down the old AOL Instant Messenger (AIM), I thought it was a huge missed opportunity for Amazon, which could have acquired AIM and then layered its overlapping micro-communities onto each customer’s profile.

Getting back to Terri, Chloe, Bean and a new SUV, if advertising moves to include interpersonalization, then on top of the picture of two women and a dog (who look a lot like Terri, Chloe, and Bean) the ad would also surface who among their friends drives a Subaru. The ad could also surface who has bought more than one Subaru.

In the early days of Facebook, Twitter and the like, advertising pundits said that social media was “word of mouth on steroids.”

Interpersonalization would be like word of mouth wearing Iron Man’s armor.
__________

 

Brad Berens is the Center’s strategic advisor and a senior research fellow. He is principal at Big Digital Idea Consulting. You can learn more about Brad at www.bradberens.com, follow him on Blue Sky and/or LinkedIn, and subscribe to his weekly newsletter (only some of his columns are syndicated here).

 

* Image Prompt: “A photorealistic, close-up picture of a target attached to a haystack. On the target, an arrow is in the second-from-the-center ring. Hanging from the arrow are a collection of seven tiny people of a variety of ages, races, genders, and clothing styles.” The LLM didn’t get the placement of the arrow right, but I like how it has the people hanging off each other. The other AIs had people all hanging from the arrow.

See all columns from the Center.

February 11, 2026