The AI Campaign Film, Between BMW's Nerve and the Coca-Cola Backlash

Within a few months, a niche technique has become a tool that can carry entire campaign films: image-to-video models like Veo, Seedance, or Runway turn a single still photo into a cinematic camera move within minutes – work that used to require a drone flight, a slider rig, and half a shoot day. BMW just demonstrated what that means with the iX3: a full, AI-generated campaign film from a major car brand, not just a small social clip to try things out. At the same time, the backlash over Coca-Cola's AI Christmas spot from last year shows how quickly the same technology can backfire.

Two brands, the same tools, completely different reactions. The difference rarely comes down to the technology itself – it comes down to what it's used for.

Why Coca-Cola earned the backlash

Criticism of AI-generated ad films runs on two levels. The first is technical: recurring patterns, slightly unconvincing motion, the now widely cited "AI slop" – audiences by now reliably recognize AI-generated moving images, especially in emotionally loaded subjects like Christmas, family, closeness. Exactly where a spot is supposed to convey warmth, technical distance stands out the most. The second level is moral: when a corporation with a well-funded marketing budget cuts costs specifically on the production that would otherwise have employed camera crews, directors, and filmmakers, that's rightly read as a signal – not innovation, but cost-cutting with a glossy finish.

I've described that exact blurriness debate in more detail elsewhere: the more brands draw on the same models, the more the look converges – and it's precisely in the subjects meant to convey closeness and humanity that this sameness stands out most uncomfortably.

Why BMW gets away with it

The iX3 spot presumably works for a simple reason: automotive advertising was already one of the most heavily technically staged genres even before image AI. Paint reflections, landscape drives, camera moves that would barely be physically possible – classic car campaigns have relied on computer graphics, compositing, and elaborate post-production for decades. An AI-generated campaign film stands out less in this genre, because the audience never assumed an unedited documentary shot in the first place. For a Christmas spot with a family by the fireplace, the expectation is an entirely different one.

That matches a distinction I've already made on campaign planning: the decisive question is never "photo, film, or AI," but what expectation a given subject triggers in the audience – and whether the chosen technique holds up to that expectation or disappoints it.

The point almost nobody considers in time

Anyone producing a full AI campaign film right now should keep an eye on a deadline that's practically absent from the current coverage of BMW and others: starting August 2, 2026, the labeling requirement from Article 50 of the EU AI Act applies to video too. A photorealistic AI campaign film that depicts vehicles, landscapes, or people true to life can fall under it – regardless of whether it's meant to deceive anyone. A campaign that goes live shortly before the deadline with no labeling concept risks not just a backlash over its visual language, but a fine on top of that for missing transparency.

What this means for our own production

In our own practice, that means: image-to-video models now belong on the tool shelf, but the decision of when to use them stays the same as with any other technique. For landscape drives, B-roll, or concept drafts ahead of the actual shoot, they save enormous amounts of time without an audience ever questioning authenticity. For subjects that depend on closeness, trust, or genuine human interaction – the cases where a campaign film comes from one hand, from concept through post-production – the camera remains the more credible choice for now. Not out of nostalgia, but because that's exactly where the audience looks closest.

Conclusion

BMW and Coca-Cola used the same technology and reaped opposite reactions, because the technology was never the actual risk. The risk lies in producing a subject with a tool that doesn't hold up to the audience's expectation – and, starting in August, in failing to disclose that in time on top of it. Whoever settles both questions beforehand can show things with AI campaign films today that used to require a multiple of the time and budget.