Google admits AI mistake in Super Bowl cheese ad

Google has corrected an inaccurate statistic about Gouda cheese that appeared in its Super Bowl advertisement featuring the Gemini AI system.

Key details: The original commercial showed Gemini generating website content that incorrectly claimed Gouda accounts for 50-60% of global cheese consumption.

• Google modified the YouTube version of the ad to remove the specific percentage, replacing it with a more general statement about Gouda being “one of the most popular cheeses in the world”
• The business owner featured in the commercial had already implemented the AI-generated content, including the incorrect statistic, on their website
• Despite Google’s correction to the advertisement, the inaccurate information remains live on the business owner’s website

Expert insight: An agricultural economics expert from Cornell University confirms that Gouda is not the world’s most consumed cheese variety.

• Professor Andrew Novakovic, E.V. Baker Professor of Agricultural Economics Emeritus, definitively refuted the original claim
• The incorrect statistic appears to have originated from Cheese.com, a website containing SEO-optimized content
• A disclaimer on the business owner’s website states that Gemini’s response is “not intended to be factual”

Corporate response: Google’s initial reaction to the error raised questions about the company’s fact-checking processes.

• Google Cloud apps president Jerry Dischler initially defended the response on X (formerly Twitter), claiming it was “grounded in the Web” and “not a hallucination”
• The error was first spotted and reported by social media user @natejhake on X
• Google subsequently chose to edit the commercial rather than continue defending the inaccurate information

Broader implications for AI content generation: This incident highlights the ongoing challenges of ensuring AI systems provide accurate information for business use.

• The situation demonstrates how AI systems can perpetuate incorrect information found on the internet, even in high-profile applications
• The presence of a disclaimer about factual accuracy raises questions about the practical utility of AI-generated content for business purposes
• This case underscores the importance of human verification when using AI-generated content, particularly for commercial applications