On Friday 6th February, 21 million viewers tuned in to watch the 2026 Winter Olympics opening ceremony. They watched Andrea Bocelli sing. They watched JD Vance get booed. And they watched an AI-generated animation so profoundly bad it united the entire internet in something rarer than Olympic gold: universal agreement.

The sequence - a ropey cartoon flashback through 100 years of Winter Olympics history, supposedly starring a strange version of White Lotus actress Sabrina Impacciatore - was, to put it charitably, uncanny. To put it less charitably, it looked like crap. Faces melted. Scenes stuttered. Limbs did things limbs don’t normally do. The IOC’s own Olympic rings were drawn incorrectly, which is like Nike getting their swoosh wrong.

“Winter Olympics using dogs**t AI slop in their opening ceremony,” wrote one viewer, in what may be the most succinct piece of creative criticism published this decade. “An animating job that a talented animator would’ve bitten your arm off for.”, wrote another. It’s strange to me that robots were used to celebrate an event all about human achievement. Maybe next time we should skip the effort altogether and just watch drones race each other instead?

Meanwhile, the BBC released “Trails Will Blaze”: a stop-motion trailer made with 700 individually 3D-printed athletes, 14 combustion techniques using real fire, a pyrotechnics team, and the BBC Symphony Orchestra performing Verdi’s Requiem. It was universally praised as a masterpiece. Same event. Same brief. Wildly different choices. Wildly different outcomes.

Two days later, Super Bowl LX delivered the second punch. Adweek reported that 23% of all ads were AI-related - 15 out of 66 national spots. Viewers dubbed it “the AI Bowl” and drew immediate comparisons to the crypto-saturated 2022 Super Bowl that preceded that industry’s spectacular collapse. Svedka’s primarily AI-generated ad scored a 7% brand match against a 63% alcohol category norm. Viewers described it predominantly as “weird” and “WTF.” I’m with them.

The Science: Humans Know Humans

Masahiro Mori’s 1970 ‘uncanny valley’ hypothesis describes the revulsion we feel when something looks almost-but-not-quite human. Our brains are exquisitely tuned face-processing machines. We detect asymmetries, unnatural textures, and odd-looking eye movement at a level far below conscious analysis. When AI-generated faces exhibit these inconsistencies, we don’t think ‘that looks AI-generated.’ We feel it. And the feeling is: ‘that was weird’.

Researchers at Bowling Green State University showed participants AI and human art without labelling the source. People couldn’t reliably tell them apart, accuracy was barely better than a coin toss. But they consistently rated human work higher on emotional engagement, self-reflection, and amusement. When asked why, they couldn’t explain it. They just knew something was off with the AI versions.

Psychologists call what’s happening the ‘effort heuristic’:  we use perceived effort as a proxy for value. The BBC’s trailer works partly because you can see the craft: real fire behaving unpredictably, visible dents on figures, light bouncing differently in every frame. These imperfections read as care. AI’s weird smoothness shows a lack of it.

Good Enough for Them?

Here’s the detail that should stop every marketing director in their tracks. When OpenAI needed to sell itself to an estimated 130 million Super Bowl viewers, did it generate the ad with its own tools? No. Its CMO was explicit: “We shot the ads with real people, on film.” Produced by Doomsday Entertainment. Human creativity. Human talent.

Anthropic’s campaign was created by Mother, directed by Jeff Low, filmed with real actors in Los Angeles. Google used live action. Meta hired Spike Lee. Every single AI company that advertised during the Super Bowl chose human-made creative to sell its product.

The one brand that did go all-in on the artificial was Svedka, using Silverside AI (the same firm behind those weird Coca-Cola AI Christmas ads) and was universally disliked by real people.

Ironically, the studio behind the BBC’s Olympics masterpiece, NOMINT, has previously produced award-winning stop-motion work for OpenAI. The company that sells AI hired a stop-motion studio to do it.

AI as the Chisel, Not the Sculptor

I use AI every day. For research and rapid iteration, it’s transformative. But there’s a critical difference between AI accelerating the creative process and AI replacing the creative product. For research, ideation, and mechanical tasks,  brilliant. Sign me up. For the final thing a human audience is meant to engage with emotionally? That’s where the Olympics, the Super Bowl, and science all show: you need humans.

I think ‘human-finished’ should become a quality standard in our industry. A genuine signal of care when many brands are showing that they don’t. Not only did the BBC make a better trailer, they made a statement about what kind of organisation they are.

In 48 hours, February 2026 delivered a masterclass. AI can get you the ad, but it can’t get you the feeling. And when we know that the point of ads is to make a human being feel something, trust the species that actually knows what feelings are. We’ve had a few million years of practice.