generative AI response featured in advertising (false fact)
In February 2025, a Google Gemini Super Bowl ad wrongly claimed that gouda accounted for 50 to 60 percent of global cheese consumption; Google removed the figure from the ad before the game aired.
Objective
Promote Gemini to small businesses during the 2025 Super Bowl by showing the AI write product descriptions, here for a cheese merchant.
The deployment
For the 2025 Super Bowl, Google runs a campaign showing Gemini helping small businesses write their content. In the spot devoted to a Wisconsin cheesemonger, the generated text claims that gouda accounts for 50 to 60 percent of global cheese consumption. The figure is false: gouda does not even appear in the top 10 according to public data. A Google executive first defends the ad by saying it is not a hallucination, since the model relied on a web source. Faced with the controversy, the figure is removed from the ad and the product page before the game airs.
Results Proof C
Established press (Fortune, TechRadar) documents the brand by name, the false figure, the executive's defense, and the correction. Google's edit of the ad confirms the facts. No financial document, hence C.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- llm Google Gemini Generative LLM staged in the campaign: it writes the cheesemonger's product description, including the false gouda statistic
- outil Ancrage sur sources web The false figure came from a web page repeated by the model; it is the lack of verification of this repeat that is at fault, not a hallucination according to Google
- plateforme Pages locales de demonstration Gemini Google product pages featuring Gemini's responses for small businesses, corrected at the same time as the spot
Post-mortem
GraveyardWhat happened sourced
In early February 2025, Google runs a Super Bowl campaign showing Gemini write descriptions for small businesses. The cheesemonger spot has the AI claim that gouda accounts for 50 to 60 percent of global cheese consumption. An observer flags the error. A Google executive first defends the ad by saying it is not a hallucination but a repeat of a web source. Faced with the controversy, Google removes the figure from the ad and the product page before the game airs.
Reason for failure sourced
The model repeated a false figure present on a web page and integrated it into validated marketing content without verification. The initial public defense (it is not a hallucination) made things worse by revealing that no factual guardrail separated the model's output from going on air. The error was all the more visible because the ad was precisely meant to tout Gemini's reliability.
Cost sourced
Reputational cost: the blunder is picked up by the tech press at the worst moment for a reliability demonstration. No direct financial cost published.
Warning signs inferred
Inferred: airing an LLM output containing a quantified statistic without human verification is a known risk. The fact that the model relies on the web does not guarantee accuracy, since the web contains errors. A factual review of the quantified claims before filming would have sufficed.
Lessons in hindsight inferred
Inferred: any AI-generated quantified claim meant to be published must go through human verification, especially in an ad that sells that AI's reliability. Distinguishing a hallucination from a repeat of a false source changes nothing for the public: it is the brand that owns the error. Checking the salient facts is not optional.
Inferred: yes, showing a generative AI write marketing content remains a valid and widespread pattern. The failure is about the lack of factual control and a clumsy crisis communication, not about the idea. With a check of the figures before airing, the same demonstration holds.
How your customers perceive this type of use
Sourced studiesUn ecart net separe les annonceurs des consommateurs : 77% des annonceurs voient l'IA positivement contre 38% des consommateurs (Yahoo/Publicis, 2024). Les mesures implicites confirment le rejet declare : en EEG, les pubs generees par IA produisent une activation memorielle plus faible que les pubs traditionnelles et sont decrites comme agacantes, ennuyeuses et confuses (NIQ, 2024). La disclosure a un effet ambivalent : elle augmente fortement la confiance quand elle est remarquee (Yahoo/Publicis), mais 27% des jeunes consommateurs disent faire moins confiance a une entreprise dont la pub est creee par IA (IAB, 2024).
Acceptance conditions
- Une disclosure visible : quand la mention IA est remarquee, la confiance globale envers l'entreprise augmente de 96% (Yahoo/Publicis 2024)
- Une qualite visuelle suffisante : les visuels IA de basse qualite augmentent l'effort cognitif et distraient du message (NIQ 2024)
Red lines
- Le contenu IA non declare puis identifie : 72% des consommateurs disent que l'IA rend l'authenticite difficile a etablir (Yahoo/Publicis 2024) et les marques utilisant des pubs IA sont plus souvent jugees inauthentiques ou non ethiques par les consommateurs que par les dirigeants (IAB 2024)
- Les mannequins et personnes generes par IA : 46% des consommateurs n'en veulent pas dans la publicite, l'inquietude premiere etant les standards de beaute irrealistes (Attest 2025)
Sources: Yahoo / Publicis Media (terrain Ebco) 2024 · IAB (avec Attest) 2024 · NIQ (NielsenIQ) 2024 · Attest 2025
How to replicate
Inference, not sourcedData prerequisites
- content to generate
- verifiable reference sources
Org prerequisites
- human factual review before publication
- fast correction process
Possible stack
- generative LLM
- verification of quantified claims
- editorial validation chain
First step: Add a mandatory human fact-check step for any generated content meant for public release, all the more so an ad that touts the AI.
The same trap, elsewhere
CNET
SEO article generation at scale
Duolingo
Public communication of an AI-first strategy
Levi's
AI-generated models and product imagery
Sources
- S1 A Super Bowl ad featuring Google's Gemini AI contained a whopper of a mistake about cheese Established press archive pending
- S2 Google fumbles Gemini Super Bowl ad's cheese statistic Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.