Peter Girnus 🦅@gothburz
I am the Chief Marketing Officer of an AI company that spent $8 million on a Super Bowl ad last night.
The ad was 30 seconds long.
It featured a child. The child asked our AI a question about the stars.
Our AI answered beautifully.
The spot tested well in focus groups.
The creative team won an internal award.
We generated 40,000 renders to find one where the AI didn't hallucinate.
The approved cut was take 39,847.
The child in the ad is not a real user.
The answer about the stars is not a real answer.
The entire commercial is a proof of concept for a product we have not shipped.
By halftime, the audience hated us.
Not just us. All of us. Every AI company that bought a spot.
Sports Illustrated reported fans were "vocally fed up" before the first quarter ended. A Harris poll conducted before the game found consumers already felt "mostly negative" about AI advertising.
We knew this. The poll was published Friday. We bought the ad in September.
I watched the game from a hospitality suite with eleven other CMOs. We had a real-time sentiment dashboard on a monitor next to the bar. By the second quarter, the needle was in the red.
Someone from our analytics team Slacked: "Audience is associating us with the Svedka ad."
The Svedka ad was an AI-generated fever dream of two dancing androids that looked like what happens when you ask a machine to render human desire without ever having experienced it. Critics called it "warm slop." We were being grouped with the warm slop.
Twenty-three percent of all Super Bowl ads were AI or tech companies, according to iSpot. More AI ads than beer ads. More AI ads than food ads. We looked at each other across the suite and realized we had made the same mistake simultaneously. We had all bought the same thirty seconds of American attention and said the same thing into it.
I watched Ring's ad from that suite. Ring built a network of millions of AI-equipped doorbell cameras pointed at American front doors. They pitched it as a lost-pet finder.
The ad said: "We built a surveillance panopticon and pointed it at your neighbors' homes, and now we're teaching it to recognize faces, but look -- a puppy." They did not use those words. They did not need to.
The audience heard it anyway. Ring accidentally described its own business model in the most damning terms possible and then asked people to feel good about it.
I watched AI[.]com's ad from that suite. Nobody in the room understood it. The domain was purchased for $70 million by the CEO of Crypto[.]com. The website crashed immediately after the ad aired, because millions of people tried to find out what they had just watched and found nothing. Seventy million dollars for a URL that could not survive its own commercial.
I watched Anthropic mock OpenAI, and then watched Sam Altman call them dishonest, and then watched two companies valued at a combined $1.35 trillion spend the rest of the evening calling each other liars on X.
Someone in the suite said: "At least we didn't do that."
I said: "We spent $8 million to say the same thing they said, and nobody noticed."
I am not sure which is worse.
By the fourth quarter, the collateral damage had started. Viewers became so hypervigilant about AI that they started accusing non-AI ads of being AI-generated. Dunkin' Donuts. Comcast. Ads made by humans, by production crews, by directors with cameras and craft services -- called fake because two hours of AI advertising had conditioned 130 million people to distrust everything on their screen.
Our industry spent $100 million on Super Bowl ads to build trust in AI.
We built the opposite.
I need to tell you about a parallel that nobody in that hospitality suite mentioned, although every person in the room was old enough to remember it.
Super Bowl XXXVI. February 2022. Crypto firms bought the ad breaks. Coinbase. FTX. Crypto[.]com. They spent $54 million collectively. The ads were flashy and confident and told 100 million Americans that the future was decentralized and inevitable and worth their money.
FTX collapsed ten months later.
Coinbase spent the following year in court.
Crypto[.]com's CEO is now spending $70 million on AI domain names.
We spent more than double what crypto spent. I know this because Tech Brew calculated it this morning and my VP of Communications forwarded it to me with no comment. She always adds a comment. The absence was the comment.
Here is what I know that I am not supposed to say.
The Super Bowl is a lagging indicator of industry health.
A lagging indicator means the peak has already happened. It means the industry already believes in itself more than the public does. It means the money has been spent, the bets have been placed, and the audience -- the 130 million people you needed to convince -- sat through your pitch and felt nothing but annoyance.
I spent $8 million to learn something that a Harris poll could have told me for free.
Nobody wants what we are selling. Not like this. Not yet. Maybe not ever. But "maybe not ever" is not a phrase that survives a board meeting, so we say "not yet" and buy another ad.
The earnings call is in six weeks. When the analyst asks about brand strategy, I will say the word "awareness" and the word "consideration" and the word "momentum."
I will not say "warm slop."
I will not say "lagging indicator."
I will not mention FTX.
I will not tell them about the sentiment dashboard, or the Slack message, or the eleven CMOs in the suite who watched the needle go red and poured another drink.
We will do this again next year.
The budget is already approved.
The budget keeps going up and to the right.
