How Google Shopping used AI to create a parade of hyperlocal ads

Deb Lee, Paul Aaron

Social Module

Share

A version of this article was originally published on Google Cloud’s blog and explores how AI is helping marketers build out-of-home creative that feels personal.

The pursuit of resonant marketing often feels like a quest for the holy grail. How can brands cut through the noise and connect with consumers on a deeply personal level, in a way that feels authentic and not intrusive? The answer might lie in hyperlocal engagement.

Last holiday season, Google Shopping and AI innovation studio Addition demonstrated the power of a hyperlocal approach with an out-of-home (OOH) campaign in New York City. This wasn’t just about advertising; it was about creating a conversation with New Yorkers in their own neighborhoods, using clever, location-specific headlines and relevant product imagery. Powered by Gemini and Imagen 3 models on Vertex AI, the campaign generated hundreds of unique ads tailored to the specific location where each one appeared. Created using local insights, data from the Maps API on Google Cloud, and relevant trends, each ad was contextually relevant to its location, right down to the city block.

An OOH ad for Google Shopping appears on a LinkNYC kiosk, a repurposed phone booth, across from the Google office in New York’s Chelsea neighborhood. The ad reads “Chelsea chic meets holiday savings” above a photo of a scarf.

If you’re a New Yorker, chances are you may have already seen these ads: clever, neighborhood-specific headlines paired with relevant product imagery. What made this campaign so innovative was its use of AI solutions at almost every step of the creative process — from identifying insights to creative ideation, understanding complex media briefs, and putting it all together. Let’s take a closer look at how generative AI and new tools, such as Imagen 3, unlock hyperrelevant marketing.

How Addition built for hyperlocal

Building for a hyperlocal campaign requires a fundamental shift in marketing strategy. It’s not just about collecting data; it’s about orchestrating all the right information. Addition designed a custom AI system built on a framework of interconnected functions. This system took in data, processed it using Gemini on Vertex AI, and put out ready-to-deploy ads using Imagen 3 on Vertex AI. Here’s a breakdown of how it all came together.

First, the team created a solid foundation. Curated and provided by the Google Marketing team, the system ingested data from three primary sources to provide context to the AI models.

  1. Media data: A CSV file containing the address and geographic coordinates of each LinkNYC kiosk where the ads were displayed.
  2. Google Maps: The system queries the Google Maps API to retrieve detailed map imagery for each location.
  3. Trend data: Public Google Trends data helped ensure the ads were featuring the season’s must-have holiday gifts.

Then they uncovered local insights and generative copy. This stage leveraged Gemini’s powerful language capabilities to generate location-specific insights. By analyzing the data inputs, Gemini identified unique characteristics of each location, from nearby landmarks to the general neighborhood vibe. These insights guided the development of clever, contextually relevant headlines.

Play animation Pause animation

Next up, the team generated images and evaluated the best ones. Google’s state-of-the-art text-to-image model, Imagen 3, generated sample product images based on the local insights and creative direction. The creative teams were then able to review and cherry-pick the final selects based on brand safety and visual consistency. This process generated a lot of options. A content evaluation model using Gemini helped cull the list to something more manageable by assessing brand safety, location relevance, and overall creative quality.

The team processed winning ads through an AI-powered video pipeline that integrated the generated text and images, optimizing the output for display on the LinkNYC kiosks.

Play animation Pause animation

A dedicated interface allowed Google’s brand managers to review the top-performing ads on a map and approve them for deployment. Once approved, the system automatically delivered the ads to the designated LinkNYC displays via a custom API.

Finally, Addition used AI to uncover real-time updates. This AI-powered system enabled a hyperlocal focus and provided flexibility for real-time content updates. During last year’s Macy’s Thanksgiving Day Parade, Addition experimented with pre- and postparade messaging, demonstrating the system’s ability to move at the speed of culture.

Two Google Shopping ads, before and after the Macy’s Thanksgiving Day Parade, appear on LinkNYC kiosks: “Parade view might be blocked, but deals are in full view,” and “The parade might be over, but holiday deals are just beginning.”

Human creative direction plus AI

The Google Shopping campaign underscores a fundamental truth. AI is not replacing human creativity; it’s augmenting it. By combining human creative direction with the Vertex AI platform and models, brands can achieve both efficiency and impact at scale.

Get started with Vertex AI, Imagen, and Gemini models today, and explore the possibilities of AI-powered creativity.

Deb Lee

Deb Lee

AI Customer Engineer

Google

Paul Aaron

Paul Aaron

Co-founder and CEO

Addition

Return to top of page