Skip to main content

(beta) Smart Audience: Use LLM powered segments to supercharge your paid strategy

Frédéric Derian avatar
Written by Frédéric Derian
Updated over 2 months ago

Smart Audience helps you identify the most relevant customer personas for each of your category, using the contextual understanding of the most advanced LLMs (like ChatGPT, Gemini, LLaMA). These segments are updated monthly, giving you fresh, actionable audience insights you can use directly in your paid media campaigns.

Instead of relying on outdated, static research, this feature generates dynamic profiles based on consumer perception, behavior, and AI-driven trends—refreshed monthly to reflect evolving customer interests.

Value Added for Clients

  • Precision Targeting: Optimize paid media campaigns with 7-12 highly detailed personas per category analysis.

  • Always Updated: Monthly updates ensure your audience data stays fresh and relevant.

  • LLM-Powered Insight: Understand how different AI models (ChatGPT, Gemini, LLaMA) perceive your ideal customers.

  • Direct Activation Ready: Persona data is structured for immediate use in advertising platforms and mapped to Google Audience Segments.


How it works ?

Smart Audience uses the power of large language models (LLMs) like ChatGPT, Gemini, and LLaMA to generate fresh, accurate personas—without you needing to run traditional surveys or panels.

1. The analysis triggers data collection via API

When you launch an analysis for a brand in a specific category, the platform request prompts to multiple LLMs, multiple times to obtain representative information.

Here is an example of prompt used :

“Based on your knowledge of online content, trends, conversations over the last 30 days but also based on pros and cons of this brand, who are the top customer profiles most likely to engage with this brand within this specific category?”

Each LLM returns its best guess at realistic and diverse customer segments, including:

  • Age range

  • Gender

  • Location type (e.g., urban or rural)

  • Household type

  • Cost sensitivity

  • Behavioral traits like:

    • What stops them from buying

    • What influences their decisions

    • What media they consume

2. Segments are aggregated & deduplicated

Because multiple LLMs are involved, some audience profiles may overlap. A second step - that combines similar segments from all LLMs into a single, unified profile and keeping the most common and useful insights - is processed.

This ensures you don’t get redundant personas, but instead a clear list of 7 to 12 strong, unique audience segments.

3. Each segment is mapped to Google "In Market Audiences"

To make activation easier for paid media teams, then we match each persona we’ve generated with the closest Google "In Market Audiences".

Based on our mapping algorithm, we define up to 5 Google "In Market Audiences" most closely aligned with our segments :

  • Those audiences are used to reach users based on their recent purchase intent from Google Ads.

  • Paid media team can also use those to create Lookalike Audiences in Meta (Facebook ads).

Several types of Google Ads campaigns available

This means paid media team gets each custom persona automatically linked to a list of pre-existing audiences in Google Ads—ready for campaign targeting or lookalike audiences.

Those audiences can be activated on multiple channels :

  • Search

  • Shopping

  • Video

  • Display

  • Gmail

  • Discovery

  • Demand Gen

4. Monthly refresh keeps you current

Markets change fast. So each month, we re-run the analysis using the same logic. This means you always have:

  • Updated personas

  • Refreshed behavior and pain points

  • Real-time Google Segment mapping

No manual surveys, no guesswork—just timely insights, driven by LLMs and mapped for paid media use.

Select from this segment to find customers who are in the market, which means that they're researching products and are actively considering buying a service or product like yours.

These segments reach consumers close to completing a purchase.


How to use it ?

1. For each Brand analysis created, a Smart Audience is available.

Go to the “Smart Audience” section. This is where you’ll find your top 7–12 customer personas.

You’ll see for each persona:

  • Who they are: age, gender, household type, town type

  • How they think: key pain points, what drives decisions, influences

  • What they use: top media channels


2. Filter by LLM Model (Optional)

Use the LLM Model Filter to compare personas generated by different models. This is especially useful if you're testing:

  • Which LLM best understands your brand

  • How different models describe buying behaviors


3. Use for Campaign Planning

Each persona is automatically mapped to the most relevant " Google In Market Audiences" so your media team can:

  • Export in csv

    • 7-12 segments are created

    • For each segment, up to 5 relevant the "In-market audiences" targeting are suggested to adapt Google targeting for Search, Display or Video.

💡 Think of each persona as a pre-built audience blueprint—ready to plug into your campaign strategies.


4. Reuse for other Use Cases

The personas and their behavioral traits can also be used for:

  • Creative briefing (tailoring tone and message)

  • Content strategy (matching formats to media habits)

  • Product messaging (addressing key pain points)

Did this answer your question?