Can AI Systems Be Fooled by Fake Brands?
A 16-month experiment by SE Ranking’s research team has shown that AI systems can be tricked into citing a fictional brand as a reliable source of information. The experiment involved creating a new fictional brand in a real niche with real competition and publishing content about it across multiple sites. The results of the experiment have significant implications for marketers and business owners who want to shape the narrative about their brand in AI search.
The experiment found that AI systems did pick up the fictional brand quickly, but most of the visibility came when the query was already connected to the brand itself. This means that new brands still need time to earn trust, build recognition, and compete for broader topics. However, when AI systems answer general industry questions, they tend to rely on established and authoritative sources.
The strongest results in the experiment came from prompts tied to information only the brand could answer, such as how the product works, how often it updates, and so on. These queries alone generated 11,430 AI answers with citations to the brand, accounting for 72% of all visibility in the experiment. This suggests that when users ask about a brand, AI systems are likely to rely on the brand’s website as one of the main sources of information.
What’s Behind the AI Systems’ Decision-Making Logic?
One of the key findings of the experiment is that AI systems do not behave alike. They vary not just in how often they mention the fictional brand, but in how quickly they pick it up, how consistently they cite it, and which domains they prefer as sources. Google AI Mode was the most reliable engine in the dataset, placing the domain in position 1 for branded queries in about 90% of cases.
Perplexity was the breakout engine for fresh content, picking up newly indexed pages within 1-3 days. However, this speed comes with a tradeoff, as Perplexity often used supporting test domains as sources instead of consistently citing pages from the main domain. ChatGPT showed the most noticeable progression over time, with visibility steadily increasing as the month progressed.
Gemini was the weakest engine in the dataset and the least consistent, struggling to identify the niche correctly and failing to include citations to the brand in about 60% of responses. These differences in behavior suggest that AI systems have different priorities and evaluation criteria when it comes to citing sources.
Who Wins and Who Loses in AI Search?
The experiment’s results have significant implications for marketers and business owners. When users ask about a brand, AI systems are likely to rely on the brand’s website as one of the main sources of information. This means that the content they cite should be fully aligned with how the brand wants to be positioned. The experiment supports this, with the “Complete Guide” page on the main site appearing in 1,799 AI answers and the “About Us” page following with 1,500 AI answers.
New brands still need time to earn trust, build recognition, and compete for broader topics. However, when AI systems answer general industry questions, they tend to rely on established and authoritative sources. This means that new brands need to focus on creating high-quality, comprehensive content that provides unique value to users.
The experiment also found that comprehensive, in-depth content earns far more AI citations than shorter articles. The strongest-performing formats were guides, reviews, and in-depth articles. This does not mean there is one ideal content length, but rather that the depth, structure, and completeness of the information provided are key factors in determining AI citations.
Steel-Man the Skeptical Case
One potential critique of the experiment is that it only tested a fictional brand in a real niche with real competition. This may not reflect the real-world scenarios where brands have existing reputations and established sources of information. However, the experiment’s results still provide valuable insights into how AI systems respond to new information and how brands can shape the narrative about themselves in AI search.
Another potential critique is that the experiment’s results may be influenced by the specific AI systems tested and the types of queries used. However, the experiment’s findings are consistent across multiple AI systems and query types, suggesting that the results are robust and generalizable.
What’s Next?
The experiment is still ongoing, and the next phase will involve testing different content formats and structures to see how they impact AI citations. The results of this experiment will provide further insights into how AI systems respond to new information and how brands can optimize their content for AI search.
One key takeaway from the experiment is that brands need to actively shape the information environment that AI systems rely on. This means creating high-quality, comprehensive content that provides unique value to users and ensuring that the content is fully aligned with the brand’s messaging and positioning.
Bookmark this one — it will matter to your business decisions this week.
By Priya Nair, AI & Startup Reporter at TrendFlashy
Ready to launch your own asset?
Check out our guide on Building a Profitable Online Business.
