PARTNER CONTRIBUTION

The development is poised to fundamentally change how digital advertising works

By Rosie O'Meara  &nbsp July 28, 2026     6-minute read

To clear up any confusion regarding marketing buzzwords, when I say, "predictive artificial intelligence (AI)-powered audience targeting," I don't mean a few lookalike audiences built using a large language model (LLM) tool. I mean targeting audiences based on what people are likely to do in the future and when they are likely to do it. Take car buying. Previously, a basic digital ad campaign to promote an auto dealership's offers on mainstream advertising platforms might have targeted an audience called "car buyers." This audience was likely created using data on prior purchase behaviors, along with recent browsing data. Not a bad approach, but it doesn't tell us everything. Are the people in this audience planning to buy this week, this month, or later this year? We don't know if they were just browsing the latest models or intending to buy a new car. We don't know if they just converted and made a purchase yesterday.

Finally, we are often limited to online activity only, bypassing a massive cohort of people who just visited a dealership. Predictive targeting changes the game. A predictive audience among car buyers will contain a group of consumers who are predicted with a high degree of confidence (85 percent or greater, according to GroundTruth parent company ZeroToOne.AI) to purchase a car within a given time.

Enabling advertisers to target based on future intent — rather than historical data alone — has led to a seismic shift in the entire advertising model. Instead of waiting for potential car buyers to visit a website and then retargeting them, predictive audience targeting effectively enables advertisers to reach people with tailored messaging the moment their decision-making process starts. The foundational model also updates continually, removing those who are no longer in-market and eliminating ad spending waste.

The benefits of predictive audiences aren't hypothetical, and initial results are astounding. Brand leaders have already reported a 60 percent increase in conversions compared to baseline, a 30 percent reduction in wasted ad spending by removing consumers that are not actually in-market, and a 27 percent hike in store visits, per the ZeroToOne.AI data.

Ultimately, the game of building audiences solely using past behaviors and hoping for the best is obsolete. The next big leap centers on prediction.

Separating the Wheat from the Chaff

What makes true predictive audience targeting different? Large behavioral models (LBMs) are the core of predictive audience targeting. Unlike LLMs, which process text, LBMs process behavioral data points to create sophisticated consumer profiles. An effective way for marketers to think about LBMs is that whereas an LLM predicts the next word in a sequence to produce a sentence, LBMs predict the next real-world behavior.

The current marketplace is flooded with "predictive AI" solutions, most of which simply aren't what they claim. Much of what passes for AI or predictive audiences today are more akin to lookalike audiences built using past behaviors and demographic data. LBM-powered predictive audiences differ significantly from these lookalikes in several ways:

• They're dynamic instead of static. A truly predictive audience updates daily with new consumers ready to take action, compared with existing models, which look backward.
• They point to actual action. A true predictive audience will tell you who is buying, what they plan to buy, and when they plan to buy it — not just who might buy.
• They are both much larger and more targeted. LBMs process and organize far more data than conventional audiences, which translates to more targetable consumers with higher purchase intent.

All three of these points can make a noticeable difference for advertisers. The dynamic nature of a true predictive audience ensures that marketers reach people as they get into purchase mode.

By pointing to actual action within specific time frames, a true predictive audience delivers enhanced results with greater efficiency. Finally, the larger audience size enables marketers to more specifically target their ad campaign and more easily scale what's working.

The core differentiator for predictive audience targeting is that it's powered by an LBM that processes billions of data points and thousands of behavioral signals at scale, updating audiences daily with the latest consumer behaviors to determine future action. The results speak for themselves.

Learning from the History of Digital Advertising

The digital advertising technology stack has matured through two eras: the targeting era and the measurement era. With the application of predictive models to digital advertising, marketers now enter a third era: prediction.

When it comes to targeting, the ad industry has been on a long and winding road. It went from choosing individual websites to programmatic selection to specific audience targeting that now enables brands to reach people based on demographics, behaviors, and location. The advertisers that outperformed were the ones who nailed down their audiences, targeted their campaigns, and kept up with the technology as it developed.

Advertisers then focused on measurement and went from impressions to clicks to conversions to measuring real-world action. Top-performing marketers spent time figuring out which metrics to optimize their campaigns for, then optimizing for full-funnel performance and creating feedback loops that turned positive outcomes into blueprints for future success.

As they enter the prediction era, marketers can learn a lot from the past winners: Early adopters who master predictive AI-powered targeting strategies will have an opportunity to pull ahead of their competitors. Using lookalike audiences will soon feel akin to buying sitcom TV ad space to reach younger audiences, instead of simply targeting ads based on demographics using connected TV.

As real-world behavioral signals move from describing past behaviors to forecasting future ones, brands that build predictive intelligence into their core data architecture will define where value creation goes, cementing their position in a new advertising landscape.

GroundTruth is a partner in the ANA Thought Leadership Partner Program.

You must be logged in to submit a comment.