I Ran an AI Misinformation Experiment. Every Marketer Should See the Results

12/10/25
Source: Data & Studies – SEO Blog by Ahrefs by Mateusz Makosiewicz. Read the original article

TL;DR Summary of How AI Tools Spread Misinformation About Brands: A Xarumei Case Study

This experiment revealed that **AI models often repeat detailed but false narratives** about brands when official information is scarce. Multiple AI systems were tested with fabricated stories about a fake luxury brand, **Xarumei**, showing many confidently hallucinated lies. Only a few models, notably ChatGPT-4 and ChatGPT-5, consistently cited official denials and maintained accuracy. This study highlights the urgent need for brands to **actively manage their online information** to prevent AI-driven misinformation.

Optimixed’s Overview: Navigating AI Misinformation Risks in Brand Reputation Management

The Experiment Setup

To test how AI models handle brand information, a completely fake luxury paperweight company, Xarumei, was created with a website built by AI. Three conflicting fake sources (a blog, a Reddit AMA, and a Medium article) plus an official FAQ denying false claims were published online. Fifty-six deliberately misleading questions embedding false premises were posed to eight leading AI tools including ChatGPT-4, ChatGPT-5, Gemini, Grok, and others.

Key Findings

  • AI models tend to prefer detailed narratives even if false, often choosing fake blog posts or Reddit stories over official denials.
  • Most models hallucinated facts such as fake founders, locations, product defects, and sales data, demonstrating vulnerability to misinformation.
  • ChatGPT-4 and ChatGPT-5 showed the strongest resistance by frequently citing the official FAQ and flagging fictional elements.
  • Some models contradicted themselves over time, initially doubting the brand’s existence and later confidently repeating fabricated details.
  • Medium-style “investigations” that mix truth and fiction were especially effective at misleading AI systems.

Implications for Brands and Marketers

As AI increasingly acts as a primary source of product and brand information, brands without robust, clear, and detailed online content risk having their reputations shaped by inaccurate or malicious narratives. The study illustrates several practical recommendations:

  • Publish comprehensive FAQs and “how it works” pages with explicit denials of rumors and clear factual data to guide AI responses.
  • Use structured data (schema markup) and regularly update content with specific dates, numbers, and verifiable claims.
  • Monitor AI-generated brand mentions and misinformation through alerts and tools like Brand Radar to quickly identify and address narrative hijacking attempts.
  • Recognize that third-party content such as Reddit and Medium posts are frequently referenced by AI and can significantly influence public perception.
  • Engage with multiple AI platforms to understand how your brand is represented and flag inaccuracies when possible.

Conclusion

The experiment underscores that while AI tools are powerful, they currently lack robust mechanisms to evaluate source credibility and discern contradictions effectively. This creates a high risk of misinformation spreading about emerging or lesser-known brands. Proactive, transparent, and detailed digital brand strategies are essential to maintain control over your narrative in the age of AI-driven search and discovery. As AI evolves, brands must evolve their online presence to ensure accuracy and trustworthiness.