Super Bowl Is Not a Developer Conference
Every few years, a technology graduates from “people like us talk about it” to “my mom texts me about it.” That is when it stops being a product trend and starts becoming culture.
This year, AI had that moment. And the stage wasn’t a keynote. It was the Super Bowl.
The Super Bowl is not a developer conference. No one is there to debate architectures. No one is comparing benchmarks or arguing about model performance. The Super Bowl is where America goes to be entertained and to argue about commercials.
And this year, AI wasn’t a background capability quietly powering someone else’s message. It was the headline.
Super Bowl LX had an unusually high concentration of AI-themed ads. Not just one or two nods to “AI-powered” features, but a meaningful cluster. Roughly 15 of 66 commercials were classified as featuring AI in some way either as the product, the plot, or the production method itself.
That last category is important.
Some brands didn’t just advertise AI. They used AI to make the ad and made that fact the story. SVEDKA’s spot was positioned as entirely AI-generated. The message wasn’t subtle: look what this technology can create now. AI wasn’t just being sold. It was part of the creative toolchain.
But what made the night especially fascinating at least if you live in product and engineering was watching the category try to solve the same strategic problem in real time: How do you make a foundational technology feel normal? You could see the different answers play out.
OpenAI framed Codex as leverage for builders. AI as a continuation of human curiosity and making. Anthropic went sharper. It leaned into satire and essentially said, “Ads are coming to AI… but not to Claude,” turning monetization into a trust battleground. When two AI companies decide the Super Bowl is the place to litigate ethics, you know the stakes are high.
Google went emotional and domestic with Gemini’s “New Home” framing. AI as co-creator in a family moment. Amazon did something clever: it played directly into the fear people already have “what if the assistant turns evil?” and then defused it with humor around Alexa+. Meta and Oakley positioned AI glasses as performance gear, blending utility with lifestyle. Salesforce injected participatory energy, using game mechanics to make AI feel interactive rather than abstract.
Then there were the category expanders.
Ring showed computer vision as a neighborhood hero helping find lost dogs. Genspark leaned into the “AI autopilots your work so you can take Monday off” fantasy, which is either delightful or deeply provocative depending on how your week is going. ai.com went minimalist and cryptic, selling inevitability more than functionality.
Across all of them, you could see the same instinct: move AI from abstraction to application. From “intelligence” to “usefulness.” That’s smart. Because usefulness is how trust begins.
But here’s the harder question.
Did these ads actually make Americans like AI more? The clearest evidence we have is behavioral. EDO’s outcomes ranking showed ai.com driving 9X the engagement of the median Super Bowl ad, measured through immediate actions like search and site visits. That is a real signal. It reflects what people did, not what critics said. But attention is not trust. People search for things they love. They also search for things they don’t understand. Or things that worry them. Or things they think are overhyped. Outcomes tell you that you created a moment. They don’t tell you whether you created confidence.
Most of the ads tried to reduce anxiety by showing tangible value. Build a home. Analyze a game. Find a dog. Automate a workflow. The strategy was clear: make AI feel concrete, helpful, human-adjacent. AI is a general-purpose technology that touches employment, privacy, creativity, information flow, and power structures.
Super Bowl LX made AI unavoidable. It put AI at the center of the cultural conversation. It even put AI inside the creative process itself. But liking AI will not be won in a 30-second spot. It will be won in the first real workflow where it either helps or doesn’t.