Ad Targeting Through Large Language Models

Tessa Burg, CTO at digital marketing agency Mod Op, lives for the “ah-ha moment” and “getting creative about capturing information” about an audience and users while respecting privacy.

The secret is using large language models (LLMs) and create a platform that "listens" for what consumers do and don’t like.

Those platforms are based on artificial intelligence and large language models, mostly using data related to how someone takes action.

The signals range from experience and intent, to historical purchase behavior, and referral sources.

Transcripts from a company’s content such as webinars or other types of YouTube-type data work well for training LLMs. This includes conversational videos and text.

Burg suggests starting with permission-based data such as past-purchase history, location and age. Go through a segmentation process and look for opportunities to collect data on personality and behavior. Frequency and duration also are important.



“Words tell you so much,” Burg says. “What people search and the frequency is very indicative of intent. You can map that out.”

This type of data has become the highest-value information, while personally identifiable information (PII) has diminished in importance.

“I don’t need to know your name and address,” Burg said. “Identifying the consumer has become the least important piece of information,” especially as more companies use LLMs and artificial intelligence (AI) to augment ad targeting.

Burg says Mod Op takes extra care to use permission-based data and to ensure the data doesn’t come from the dark web.

When Google released Performance Max, for example, Mod Op created tools in its marketing tech stack to adjust the way it leverages core algorithms to deliver ads across the channels.

“I used to determine channels and adjust bids, but now I’m looking at what are the best messages and creatives for an audience,” she said. “The platform now determines where it gets served and how much I should spend. Then you can use the performance data to personalize in real time.”

She says some of the most transformative AI tools in marketing today include conversational AI platforms such as ChatGPT and now Llama 3, predictive analytics, and AI-driven content generation tools.

“These tools enhance creative processes by automating routine tasks, providing data-driven insights, and enabling personalization at scale,” she says.

For instance, conversational AI can serve as a strategic assistant, offering new ways to engage customers directly and gather insights, while AI-driven content tools can generate a variety of marketing materials quickly, helping brands maintain a consistent voice across all channels.

On the analysis side, these platforms can consume a ton of data and identify patterns and extract insights in seconds and pull it into a visual dashboard.

Mod Op has been on an acquisition spree during the past five years to build out its services, acquiring at least eight companies. Four of those occurred in 2023. In October 2023, it announced the acquisition of Crenshaw Communications.

Next story loading loading..