AI in Super Bowl LX Ads: Who Used It and How

Super Bowl LX in February 2026 was the first Big Game where AI was both the biggest advertiser and the biggest production tool, and the two facts are worth separating. On one side, tech companies bought slots to sell AI products: OpenAI, Anthropic, Meta, Amazon, Google, Wix. On the other, a large share of the night's advertisers quietly used generative AI somewhere in the production pipeline without saying so. Industry estimates put generative AI in the production of 50% or more of the night's spots, mostly in pre-production and mostly undisclosed. This is the story of who used AI on the record, who used it off the record, and what an $8-million-per-30-seconds ad break tells you about where production economics are heading.

TL;DR

• AI was the single largest ad theme in Super Bowl history. Reporting identified roughly 15 of 66 spots (about 23%) as AI-focused, either selling an AI product or produced with AI.
• Svedka aired what it called the "first primarily AI-generated" national Super Bowl spot, a robot-party ad that took about four months to build with AI studio Silverside.
• Anthropic's Claude ran ads mocking AI advertising and started a public feud with OpenAI's Sam Altman. Those same films later won two Film Grand Prix at Cannes.
• Most production AI was invisible, used in pre-production and never disclosed, which is the real signal for brand teams.
• The cost context is the point: a 30-second slot ran around $8 million, while the creative inside it is now the cheapest part of the equation.

The AI product parade

A striking share of the night's inventory was AI companies advertising to consumers. OpenAI ran "You Can Just Build Things," centered on ChatGPT and its Codex tool as everyday creative enablers. Anthropic made its first Super Bowl appearance with two darkly comic films for Claude, built around the idea of an AI assistant interrupted by intrusive ads, a jab at the prospect of advertising inside chatbots. Meta showcased its Oakley AI glasses with slow-motion capture and hands-free posting, fronted by IShowSpeed and Spike Lee. Amazon ran a satirical Alexa+ spot with Chris Hemsworth. Google highlighted image generation with Nano Banana Pro for home design. Wix launched Wix Harmony, a hybrid conversational-plus-drag-and-drop website builder. Ring, Hims & Hers, Ramp, and Rippling rounded out the AI-feature spots.

The Anthropic entry generated the most heat. Its critique of ads-in-chatbots prompted OpenAI's Sam Altman to publicly call the ad "clearly dishonest," turning into what observers described as a nerdy, corporate version of a celebrity feud. It also, notably, worked as advertising: the campaign went on to win two Film Grand Prix at Cannes Lions 2026.

Svedka: the on-the-record production milestone

The vodka brand Svedka aired what it described as the first "primarily" AI-generated national Super Bowl spot, a 30-second ad featuring robot characters dancing at a party. The production is more instructive than the ad itself. It took about four months to reconstruct the characters and train the AI to replicate consistent facial expressions and body movement, in partnership with AI studio Silverside. Humans drove the storyline.

Two things stand out. First, "primarily AI-generated" still meant four months of work and a specialist studio, not a weekend and a prompt. The character-consistency problem, keeping the same robot recognizable across shots, is exactly the constraint that separates a demo from a broadcast-ready spot. Second, the choice of robots as the subject is not incidental. Like Kalshi's chaotic characters before it, a robot party is a register where AI's stylized, slightly synthetic quality reads as intentional rather than broken. That is a craft decision, and it is the same one every successful AI spot has made.

The invisible majority

The more important number is the undisclosed one. Estimates put generative AI somewhere in the production of half or more of the night's advertising, concentrated in pre-production: storyboards, mood work, previsualization, concept exploration, background plates, and cleanup. Almost none of it was disclosed, because at that layer there is nothing controversial to disclose. AI used to plan a shoot is not the same as AI used to replace one.

This is the actual trend for brand teams. The visible AI ads are a story about the AI industry buying attention. The invisible AI is a story about how commercials get made now. The storyboard and previsualization stages have moved to AI across a wide swath of the industry, silently, because it is faster and cheaper and produces nothing an audience ever sees as "AI." The disclosure debate is loud about the 10% that ends up on screen and quiet about the 90% that shapes the shoot.

The cost context nobody says out loud

A 30-second Super Bowl slot ran in the neighborhood of $8 million for LX. Against that media cost, the creative is a rounding error, which is exactly why the AI production story cuts differently at this altitude than it does in performance marketing.

When the media buy is $8 million, saving 90% on production does not change the economics of the ad, it changes who can afford to make a good-looking spot at all. Svedka and Kalshi-style productions mean a challenger brand with a media budget but no eight-figure production war chest can now put broadcast-quality creative on air. The leverage is not on the Super Bowl advertiser who was always going to spend heavily. It is on everyone one tier down, the regional and mid-market brands whose entire annual video budget used to be less than one Super Bowl production. For them, the collapse in production cost is the whole game.

What this means for brand teams

Judge the tool by the register, not the hype. The AI spots that landed, Svedka's robots, Claude's comedy, chose subjects and tones where the technology's fingerprints were an asset or invisible. The ones that fail try to pass generated footage as a warm, handmade, human moment. Match your use of AI to a register that flatters it.

The real adoption is in pre-production. If you are waiting for permission to run a fully AI final spot, you are watching the wrong layer. Move your storyboard, previz, and concept exploration to AI now. That is where the industry already went, and it carries almost none of the disclosure or brand-safety risk of on-screen generation.

Disclose the on-screen stuff, not the process. The norm settling across the industry is that AI used to make what the audience sees may warrant a note, while AI used to plan a conventional shoot does not. Keep an honest internal log either way. The rules on disclosing AI ads are still hardening, and the safe posture is transparency about output.

The Super Bowl is a spotlight, not a representative sample. What LX actually showed is that AI has become normal infrastructure in commercial production while remaining a novelty on screen. The brands that win the next few years will be the ones treating it as infrastructure. If you want a canvas where the pre-production layer, storyboards, previz, concept frames, variant tests, lives in one place across every major model, start on 8frame.

For the campaigns that earned their place beyond the Big Game, see the best AI-made ads of 2026. For how these spots get judged when they enter awards, read how award juries judge AI work in 2026.