Last winter, a 23-year-old in Nashville named Jessica Rizzardi posted a dance on TikTok, and a vodka brand taught two robots to do it on the biggest ad stage in America. That routine was the most human thing in Svedka’s Super Bowl LX spot. Almost everything around it was rendered.
Nine months later, Svedka is offering $500,000 to anyone who can make it a better 15-second ad. Entries close December 1. Andy Cohen, who will help pick the winner, launched the contest by interviewing the brand’s robot mascot on YouTube and calling the last spot “bold, bizarre and a dystopian nightmare.” That is the face of the contest describing the thing the contest exists to replace.
We want to be careful about what happened here, because the lazy read is “AI bad, humans good,” and that read is wrong in a way that is going to cost a lot of people a lot of money.
Svedka didn’t stumble because it used AI. The new contest toolkit openly welcomes AI-made entries. The spot stumbled because, by its own industry’s account, there was no idea in it. Adweek ran its creative-reaction piece under a headline arguing that the ad was forgettable and that AI is not an idea. The machine executed flawlessly. It just had nothing worth executing.
Now look at what the brand did next. It didn’t fire the robot. It didn’t swear off the tools. It put half a million dollars on the table and asked strangers for taste. A brand owned by Sazerac, one that Nielsen data cited by the company calls the top volume growth driver in U.S. spirits over the past year, admitted in public, with a prize purse, that the scarce input was never rendering power.
It was judgment.
Somebody deciding what’s funny, what’s true and what deserves fifteen seconds of a stranger’s attention. That decision is the whole job, and for about a decade the industry has been trying to convince itself that the job was the file.
Coca-Cola has now run this experiment twice. In November 2025 it released its second consecutive AI-generated “Holidays Are Coming” campaign, built with outside AI studios. Global CMO Manolo Arroyo said AI cut the production timeline from roughly a year to a month, and the teams behind it generated more than 70,000 video clips in 30 days. The reaction from artists, animators and a large slice of the public was that the most sentimental ad on the calendar had somehow lost its heart.
Seventy thousand clips. One verdict.
That is the math nobody selling pure efficiency wants to sit with. When production approaches free, the volume of stuff explodes and the value of any single piece of stuff collapses. You can generate a thousand variations of a fembot in a nightclub before lunch. You cannot generate the reason anyone should care about the fembot, the nightclub or the vodka.
Consumers are already pricing this in, and they are not subtle about it. In October, Auros and Talker Research surveyed 4,000 adults across the U.S., the U.K. and Australia. Forty-four percent of Americans said they go out of their way to do business with brands that don’t use AI in their advertising. Forty-three percent said they had felt duped by a product advertised with AI. Fifty-three percent said they don’t trust AI-advertised products to be described accurately.
That isn’t a quirky tech preference. That’s a trust tax, and it gets collected at checkout.
Put that next to the stakes Svedka chose. Its spot was billed as the first national Super Bowl commercial devoted solely to vodka in more than three decades, and it brought back a mascot after a twelve-year break. A historic slot, a revived icon, a brand on a growth tear, and the most memorable thing about the night became the argument over how it was made.
So production got cheap. Taste didn’t. If anything, taste got more expensive, because now it is the only thing separating your brand from the infinite scroll of competent, frictionless, forgettable content that every tool on earth can produce on demand.
Here is where it gets uncomfortable for our side of the table.
The obvious counterargument writes itself, and we’ve heard it in pitch rooms: if the audience can do better, why pay anyone? A kid in Nashville choreographed a Super Bowl spot from her phone. A contest can source the next one for $500,000, which is a rounding error against a national media buy. Maybe the creative class is the thing being disrupted here, not the robot. Maybe the confession is ours.
It’s a fair hit. It’s also wrong, and the contest itself proves it.
Look at what Svedka is actually buying. Not a file. A prize buys a funnel: thousands of submissions, a celebrity judge with an eye, a brand team that will have to choose one and defend it to a boardroom. The contest is a taste-outsourcing machine, and it still ends with a human picking. The $500,000 is the price of admitting that picking is the job. Crowdsourcing doesn’t eliminate creative direction. It shoves it to the very last step and hopes that’s enough.
It usually isn’t, because the brief decides what comes in. Ask for “the funniest 15 seconds about vodka water” and you’ll get ten thousand people doing the same three jokes, shot in the same kitchen, with the same ring light. You cannot judge your way out of a bad question. The brands that are winning right now wrote a better question before anyone opened a generator or a camera app.
And this is where the culture part stops being decoration. Walk down Fairfax on a Saturday and watch the line outside a drop. Nobody is standing there because the product photography was rendered well. They’re there because somebody with taste decided what the drop would be, who would wear it first, which corner to shoot it on and which kid in the neighborhood got to post it before the brand did. Every person in that line can tell the difference between something made by people who give a damn and something generated by people who didn’t. A decade of feeds trained them to smell it in under a second.
That instinct is what Svedka is really paying for. The premise of the contest is that the audience believes it could have done better. They aren’t bragging about their editing software. They’re saying: we know what’s good, and you shipped something that wasn’t.
Believe them.
For the record, we are not a shop that clutches its pearls about the tools. We use AI every week for storyboards, shot lists, versioning, rough cuts and the kind of cleanup that used to eat a junior editor’s entire Thursday. It is a phenomenal intern. But the work that actually moves people still starts in a room where somebody says “no, not that” forty times in a row, until the forty-first idea is the one that feels true to the brand and alive in the culture it’s entering.
That “no” is the most expensive thing we sell. It’s also the one thing the rendering engine can’t produce, because it requires knowing the audience well enough to predict what will bore them, what will offend them and what will make them send it to the group chat at 1 a.m.
So here’s the honest read on Svedka’s $500,000. It isn’t a stunt and it isn’t a surrender. It’s a receipt. It’s what a brand pays after learning, on the most expensive night in advertising, that taste was always the line item and the robot was always the cheap part.
Smart brands will learn it from Svedka’s receipt instead of their own.
The robot can dance. It just can’t tell you whether the dance is any good.
