Lidl
participatory genAI campaign: UGC image generator in the brand palette
In 2025, Lidl France and the agency Marcel (Publicis) launched Lidlize, a genAI app (fine-tuned Bria model plus OpenAI and spaCy) that generated more than 1.7 million brand visuals in three weeks, with more than one million participants and peaks beyond 1,000 image requests per minute.
Key points
- Lidlize genAI app letting the public create objects in the Lidl brand palette.
- Fine-tuned Bria model, OpenAI LLM for intent, spaCy for moderation, on AWS.
- 1.7 million visuals and more than 1 million participants in 3 weeks.
- Evidence level B, status mixed signals: one-off campaign awarded at D&AD 2025.
Objective
Bring the Lidl brand to life with the general public by letting them create objects in the retailer's visual identity themselves, with a viral mechanic: the most liked creation is actually made and sold.
The deployment
Lidl France, with the agency Marcel (Publicis), launched an app called Lidlize that lets anyone generate a visual in the brand's red, blue, and yellow palette. The user types the name of an object, a car, sneakers, a pet figurine, without any prompt engineering skill. Behind it, an OpenAI LLM interprets the intent and composes a structured prompt, spaCy applies the allowed-content rules, and a Bria model trained on Lidl's visual language produces the image so that it stays recognizable as Lidl whatever the requested object. The creations were displayed in a public gallery where the community voted: the most liked was to be actually made and sold by the retailer, which provided the incentive mechanic. The system runs on AWS, with multilingual input (French entries translated) and a generation time cut from 8 to 2 seconds to handle the load.
The case in action
Press coverageLidl - Lidlize (case study) · voir sur YouTube
Results Proof B
Quantified case study from the vendor Bria, plus established press (PYMNTS) naming Lidl and Marcel, plus an official D&AD archive entry confirming the scale of participation. Concordant sources on the same orders of magnitude. Not level A: no financial result or attributed revenue.
How it works
Documented architectureThe stack in detail
- llm Bria Image generation model fine-tuned on Lidl's visual language
- llm OpenAI Interpretation of user intent and generation of the structured prompt
- outil spaCy Open-source NLP for applying the allowed-content rules
- infra AWS Hosting and auto-scaling to handle the load peaks
How it runs, concretely
For ops teams-
1Free user entry client / general public
The user types the name of an object in the app, without prompt engineering, in French.
-
2Prompt interpretation and structuring AI
An OpenAI LLM understands the intent and composes a standardized structured prompt; spaCy filters the allowed content.
-
3Generation in the Lidl visual language AI
The fine-tuned Bria model produces the image in the brand's palette and style, in about 2 seconds.
-
4Display and community voting client / brand team
The creation enters a public gallery where the community votes; the most liked visual is meant to be made and sold.
The user's intent (the word or object entered): this is what the LLM turns into a structured prompt. If intent interpretation or moderation fails, off-guideline or forbidden content is generated, and the image no longer looks like Lidl.
How your customers perceive this type of use
Sourced studiesUn ecart net separe les annonceurs des consommateurs : 77% des annonceurs voient l'IA positivement contre 38% des consommateurs (Yahoo/Publicis, 2024). Les mesures implicites confirment le rejet declare : en EEG, les pubs generees par IA produisent une activation memorielle plus faible que les pubs traditionnelles et sont decrites comme agacantes, ennuyeuses et confuses (NIQ, 2024). La disclosure a un effet ambivalent : elle augmente fortement la confiance quand elle est remarquee (Yahoo/Publicis), mais 27% des jeunes consommateurs disent faire moins confiance a une entreprise dont la pub est creee par IA (IAB, 2024).
Acceptance conditions
- Une disclosure visible : quand la mention IA est remarquee, la confiance globale envers l'entreprise augmente de 96% (Yahoo/Publicis 2024)
- Une qualite visuelle suffisante : les visuels IA de basse qualite augmentent l'effort cognitif et distraient du message (NIQ 2024)
Red lines
- Le contenu IA non declare puis identifie : 72% des consommateurs disent que l'IA rend l'authenticite difficile a etablir (Yahoo/Publicis 2024) et les marques utilisant des pubs IA sont plus souvent jugees inauthentiques ou non ethiques par les consommateurs que par les dirigeants (IAB 2024)
- Les mannequins et personnes generes par IA : 46% des consommateurs n'en veulent pas dans la publicite, l'inquietude premiere etant les standards de beaute irrealistes (Attest 2025)
Sources: Yahoo / Publicis Media (terrain Ebco) 2024 · IAB (avec Attest) 2024 · NIQ (NielsenIQ) 2024 · Attest 2025
How to replicate
Inference, not sourcedData prerequisites
- A brand visual language codified enough (palette, shapes, style) to fine-tune a generation model
- Licensed training datasets, to avoid the risk of unsourced content
- Explicit allowed-content rules to translate into moderation filters
Org prerequisites
- A creative agency or team able to orchestrate the intent LLM, moderation, and the image model
- A real incentive mechanic on the retailer side (here, making the winning product) to trigger virality
- An infrastructure able to absorb sudden traffic peaks (auto-scaling)
- A legal review of submitted and generated images (GDPR, image rights, AI Act)
Possible stack
- Image generation model fine-tunable on brand guidelines (like Bria)
- LLM to turn a free entry into a structured prompt
- NLP library for moderation (like spaCy)
- Cloud with auto-scaling
- Web gallery with a voting system
First step: Verify that you can fine-tune a generation model on the brand guidelines with licensed data: this is what makes each image recognizable and legally defensible.
The same pattern, proven in other industries
Adobe
in-house genAI production of marketing content at scale
Atlassian
internal genAI agents for account targeting and marketing content production
BMW
advertising and experiential creation with generative AI
Sources
- S1 How Lidl Used GenAI to Fuel a Viral Grocery Marketing Campaign Established press archive pending
- S2 Lidlize - D&AD Awards archive (Wood Pencil, Creator Content, 2025) Primary archive pending
- S3 Case Study: Personalized AI Marketing Campaign - Lidl (Bria) Interested party archive pending
- S4 How Lidl Went Viral with 1.7 Million AI-Generated Products Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.