- Lidl’s GenAI app “Lidlize” generated over 2 million images in three weeks, with a peak of 1,000 image requests per minute.
- The platform removed friction by eliminating prompt engineering, using LLM logic to auto-generate structured prompts.
- A custom AI model trained on licensed datasets ensured brand fidelity and copyright compliance, supporting the EU AI Act.
- The campaign’s viral mechanic? The most-liked creation would become a real-life product sold by Lidl.
- This activation proves that when co-creation meets compliance, GenAI can drive mass engagement, even for groceries.
The grocery chain handed over the brand to the people, and the internet turned shopping baskets into cat-shaped waffle makers, all in Lidl’s signature colors.
When Lidl France launched Lidlize, a generative AI-powered platform developed in collaboration with creative agency Marcel Worldwide and AI startup Bria, it wasn’t just a flashy gimmick. It was a bold, brand-first experiment in democratizing creativity, inviting users to co-create marketing content and blurring the line between fan and brand owner.
The results? Over 1.7 million AI-generated images in just three weeks, with peak activity reaching 1,000 image requests per minute.
What started as a playful tool turned into a cultural moment, powered by responsible GenAI design and a clear, viral incentive: the most liked design would become an actual Lidl product.
GenAI for the Masses, Not Just the Tech-Savvy
At the heart of Lidlize was simplicity. Unlike most generative AI tools, Lidlize didn’t require users to understand prompt engineering. Instead, users simply typed in any object—say, “toaster” or “handbag”—and the tool returned a fully formed Lidl-themed visual, wrapped in the brand’s iconic red, blue, and yellow palette.
This seamless user experience was no accident. Marcel built a system that automated prompt generation using large language models (LLMs), translating free-form input into structured requests behind the scenes. This lowered the barrier to entry, making the campaign accessible to casual users, not just AI enthusiasts or designers.
The experience felt more like a toy than a tool—but that’s precisely what drove its virality. People weren’t just testing tech—they were playing with the brand, and by extension, showcasing their creations across social media. The campaign spread organically, amplified by influencers and fans who saw their Lidlized creations as digital collectibles.
An AI Engine Built for Scale, Speed, and Ethics
While the frontend focused on fun, the backend tackled serious challenges. Bria’s AI model was trained on fully licensed, commercially safe datasets, which meant every generated image was legally above board, critical in a regulatory environment increasingly focused on AI governance, especially under the EU AI Act.
The system also featured content moderation safeguards, including semantic filtering powered by OpenAI models and NLP tools like spaCy. This ensured inappropriate or off-brand requests were automatically flagged or transformed, maintaining both ethical integrity and brand consistency.
@lidlfrance 🔥 On débarque sur TikTok 🔥 Pour l’occasion, on vous a concocté une DINGUERIE ! Faites vos propositions et votez sur www.lidlize.com. Les créateurs des 3 objets les plus likés seront récompensés 🎁 On attend vos créations avec le #Lidlize ♬ son original - Lidl France
Critically, Marcel’s infrastructure design meant that the platform could scale automatically with demand. Image generation time was optimized down from 8 seconds to just 2, a huge differentiator when users expect instant gratification, and when you’re processing thousands of requests per minute.
At the height of the campaign, this efficiency kept things running smoothly despite unexpected viral traction.
Localization Meets Brand Control
One of the lesser-known but essential features of Lidlize was its multilingual support. French-language inputs were automatically translated into English behind the scenes, ensuring the AI model could process them accurately. This localization effort allowed Lidl to stay true to its French customer base while leveraging global AI tools.
Meanwhile, the brand’s visual identity remained intact. Bria’s model was fine-tuned on Lidl’s specific design language, ensuring that no matter what object was imagined, the final image felt distinctly Lidl. From fashion accessories to food items, every output upheld a consistent, recognizable aesthetic—a challenge that many GenAI campaigns fail to manage.
This fusion of open creativity with brand control created a safe sandbox for user-generated content. Fans could remix the brand without distorting it, giving Lidl a massive trove of user engagement without diluting its image.
More Than a Gimmick: What This Campaign Tells Us
Lidlize wasn’t just a moment of internet novelty—it was a proof-of-concept for how generative AI can drive meaningful, participatory marketing at scale. And it worked because every aspect of the campaign was thoughtfully engineered:
- Frictionless UX: No login walls, no complex prompts, just type-and-play creation.
- Built-in virality: A clear incentive (the winning item gets made IRL) encouraged sharing and engagement.
- Regulatory foresight: Every asset was traceable, licensed, and compliant with AI laws.
- Scalable infrastructure: No crashes, no slowdowns—just high-speed generation in real time.
Most importantly, it showed that GenAI doesn’t have to be locked behind technical jargon or enterprise dashboards. With the right guardrails, it can be a playground for everyday consumers—and a powerful brand amplifier for businesses.
