This year, AI firms flocked to buy Super Bowl commercials, hoping to make a huge splash about their new products in a cost-effective way. A 30-second spot during Super Bowl LIX cost about $8 million, which may seem like a lot to pay for a short ad. Given that the massively popular game drew in more than 200 million viewers, however, a commercial’s cost-per-viewer actually amounts to less than a nickel. It’s incentives like these that motivated Google to use the Super Bowl to promote Gemini, its AI writing assistant, to business owners. But when viewers spotted a factual error in a Gemini commercial just days before the big game, it raised concerns about how much generative AI technology has actually advanced.
In the ad, a cheese-maker in Wisconsin asks Google’s AI tool for help writing a description to put on their website for Smoked Gouda cheese. Gemini responds with a write-up that states Gouda makes up “50 to 60 percent” of all the cheese eaten in the entire world. This is far from true. Experts say that cheddar is the most popular cheese in the world, with mozzarella and Parmesan also counting for a significant percentage. The error raised an important question about where the incorrect statistic came from: Was Gemini “hallucinating,” a term for when AI systems invent untrue information?
After the error was discovered, Google said its AI tool did not hallucinate and the mistake was caused by Gemini pulling false information from an online source. The incorrect numbers Gemini used can be found on several websites, including the Wisconsin Cheese Mart website and Cheese.com, without citing any original source. Google corrected the ad before it aired during the Super Bowl, with Gemini calling Gouda “one of the most popular cheeses in the world” without giving specific numbers. But the incident reveals how experimental AI technology still is. Even in Super Bowl ads, which get extra scrutiny because of the game’s extremely high profile, AI results still need a fact check.
Questions:
- Why did many AI companies choose to advertise during this year’s Super Bowl? Do you think they were successful?
- What can Google’s incorrect Gemini advertisement potentially tell us about the effectiveness of AI?
Sources: Graham Fraser and Tom Singleton, “Google Remakes Super Bowl Ad After AI Cheese Gaffe,” BBC, February 6, 2025. Kyle Orland, “Not Gouda-nough: Google Removes AI-Generated Cheese Error From Super Bowl Ad,” Ars Technica, February 5, 2025.