France’s PMU says 100% of its digital ad creative now runs through AI—here’s what that really means

PMU, France’s state-backed horse-racing betting operator, says it has shifted to a workflow in which 100% of its digital marketing assets are produced using AI—an unusually sweeping claim as brands experiment with generative tools but rarely say they’ve fully scaled them.

The details came from Olivier Pribile, PMU’s marketing director, speaking July 1 at the Directeurs Marketing Summit 2026, an event organized by French trade publication Stratégies and hosted at Webedia Elephant. PMU framed the move as a gradual ramp-up over roughly the past year and a half, culminating in AI being used across the full range of online ad formats the brand runs.

Behind the headline number is a more practical question: which parts of “production” are automated, and which still require human hands—especially when the creative has to look real enough to protect brand trust.

PMU’s marketing chief makes the “100% AI assets” claim at an industry summit

On July 1, Pribile told the summit audience that within the scope of PMU’s digital assets, 100% of production is done via AI. In context, PMU is talking about creative built for digital channels—banners, social media formats, activation variations, and content adapted to different placements.

PMU described the shift as an industrialized practice rather than a pilot. The company says it began about a year and a half ago without aiming for a dramatic break from past methods—more a learning curve than a forced transformation.

Pribile used a concrete example to explain the early hurdles of getting AI outputs to match the brand’s world of racetracks and horses. “Getting creative with horses running on racetracks while saying there’s a 100 bonus at sign-up, you’d think it’s simple, but at the start, it’s more a pony that the AI gives us than a real horse,” he said.

The setting—Webedia Elephant’s offices—and the audience of marketing leaders also made the talk a kind of proof point: PMU positioning itself as a consumer brand that has moved beyond AI as a novelty and into AI as a standard part of the creative workflow.

Even so, the “100%” line needs careful reading. In marketing, “digital assets” can mean finished creatives ready to run, but also variants, graphic elements, short animations, and associated text. As described, PMU’s claim points to AI being used in production, not necessarily AI operating without human oversight. In practice, teams still define prompts, select outputs, retouch, check compliance, and approve final materials.

That distinction matters because the real test isn’t man-versus-machine—it’s whether a mass-market brand can keep quality consistent. One visually off campaign can quickly become a punchline or trigger distrust on social media.

Why PMU is leaning on AI: more versions, faster turnaround, more formats

PMU’s operational pitch is straightforward: produce more content, faster, across more formats. Digital platforms increasingly demand fragmentation—different image ratios, short video durations, A/B iterations, and creative tailored to specific placements.

In that environment, AI becomes a “variation engine.” PMU says it has generalized AI use across all its digital formats—typically covering activation, display, social, CRM assets, landing pages, and campaign graphic elements.

The article also cites other public remarks attributed to Pribile referencing a threshold of 90% of digital formats produced with AI assistance, along with a 30% increase in production. The coexistence of 90% and 100% may reflect different scopes or rapid change between communications. The core point remains the same: PMU is describing near-total coverage of the relevant formats and a faster creative cadence without a proportional increase in headcount.

But realism remains a make-or-break constraint. Generative models can produce anatomically incorrect animals, inconsistent textures, or impossible perspectives—errors that stand out immediately in advertising and can damage credibility. The “pony” anecdote signals the kind of work required: more precise instructions, model choices, iteration, and tighter selection.

There’s also brand consistency. PMU isn’t just selling a promotion; it’s selling an identity tied to horse racing and a regulated offering. AI-generated scenes still have to follow a brand rulebook—colors, typography, tone, and how racing is depicted. In that framing, AI becomes a tool for composition and controlled variation rather than a free-running generator.

Speed only matters if it improves marketing control. More versions allow teams to test hooks, images, and bonus mechanics and then optimize campaigns. PMU’s approach is presented as part of that performance-driven loop—keeping pace with activations without long production cycles.

Quality control, rights, and compliance: the guardrails behind AI visuals

A claim of 100% AI-produced digital assets immediately raises governance questions. For a consumer-facing brand, legal and compliance issues aren’t a technical footnote. First, PMU has to avoid generating misleading content. When an ad mentions an offer, a bonus, or conditions, the image and copy must align with reality and required legal disclosures. AI can speed production, but it doesn’t automate responsibility.

Second is rights management. Generative AI tools rely on models trained on large datasets, and debates over training data and copyright remain active. At the advertiser level, the cautious approach is to define tool usage terms, document sources and settings, and avoid prompts that too closely reproduce an identifiable style or protected element. In image-heavy industries, the risk of claims exists even as case law evolves.

Third is brand protection. AI visuals can introduce unwanted details—logo-like elements, signage, nonsensical text, or inconsistent inscriptions. For official communications, those artifacts are a problem. That pushes teams toward systematic quality control: checklists, detail verification, animal anatomy checks, and internal-rule compliance. In some cases, the review burden can be heavier than in traditional production because errors can be subtle.

Fourth is internal transparency and traceability. Saying all digital creation runs through AI implies shared tools, procedures, asset storage, and at least minimal tracking. In larger organizations, that also touches cybersecurity—avoiding sensitive information in external tools, anticipating leakage risks, and controlling access.

Finally, there’s an ethics and representation dimension. Even without deepfakes of real people, AI can generate stereotyped or unrealistic scenes. In racing and gambling, advertising is scrutinized for responsibility and audience protection. Scaling AI means building those criteria into production, not bolting them on at the end.

What PMU’s move signals for in-house creative work

PMU’s announcement lands as marketing departments weigh the mix of in-house production, agencies, studios, and AI tools. Saying 100% of digital assets are made with AI doesn’t automatically mean creatives disappear. It can mean the work shifts—less manual production, more direction, selection, retouching, test strategy, and performance-driven iteration.

In a traditional model, a digital campaign runs through a brief, a creative phase, and then often-heavy adaptation work. With AI, some adaptations can be generated in series, changing the relationship to time. Creative can move closer to performance management, with variants produced continuously based on results. That can shift pressure onto teams to deliver quickly while maintaining consistent quality.

The horse-and-racetrack challenge illustrates a broader point: brands with a specific visual universe have to learn how to guide or constrain models to get reliable outputs. That can mean internal references, libraries of approved images, standardized prompts, specialized models, or hybrid workflows combining generation and retouching. PMU’s “horse versus pony” gap is presented as part of that learning process.

For agencies and outside partners, the signal is mixed. Some volume production may be internalized or accelerated, while other work becomes more strategic—brand territory, storytelling, identity, major campaigns, and guidance on legal frameworks and governance. The market may split between high-volume production that AI makes easier and premium creative where singularity and control remain paramount.

PMU’s communication also raises a measurement question. If the company is claiming full coverage and higher output, the next debate is about metrics—average turnaround time, cost per asset, reuse rates, media performance of variations, and error rates caught in quality control. Those numbers weren’t detailed in the reported remarks, but they typically shape what “scaling” really means. For other advertisers, PMU’s case serves less as a universal model than as evidence that broad deployment is possible—if the methods and guardrails are built first.

Frequently asked questions

What does “100% of digital assets are made with AI” mean for PMU? It refers to creations intended for digital channels—banners, social variations, activation formats—produced using AI tools within the workflow. It does not mean human teams disappear, since selection, retouching, brand approval, and compliance remain handled by professionals.

Why are horse and racetrack visuals especially challenging for AI? Generative models can produce anatomical errors, inconsistent perspectives, or unrealistic details. In a highly recognizable world like racing, those flaws are easy to spot and can hurt credibility, requiring iteration, quality control, and art-direction constraints.

What are the main risks of heavily automated creative production? The risks center on visual quality, brand consistency, ad-message compliance, rights management, and unwanted artifacts. Governance, traceability, and data protection also come into play.

Does AI replace creative jobs in this kind of organization? It mainly reshapes them. Manual production may decline in favor of AI-oriented art direction, prompt design, quality control, rapid variations, and performance-led management. Skills shift toward supervision and validation.

Key takeaways

PMU says 100% of its digital assets now run through AI in 2026, after a gradual ramp-up over about a year and a half. Realistic racing visuals remain a key quality checkpoint—captured by Pribile’s “pony” example—while the broader goal is higher-volume, faster production with compliance and brand guardrails.

Sources

La Revue du Digital; “Directeur marketing du PMU : « 100% de nos assets digitaux sont faits en IA »” (La Revue du Digital); Stratégies; CB News; Dailymotion (“Cyrille Giraudat, dans 01Business – 28/06 1/4”).