AI models are far more likely to search for brands they already recognize, according to new research. In nearly two-thirds of brand-specific queries, the top five familiar brands were favored. The findings raise questions about fairness in AI-driven discovery.
Artificial intelligence models are significantly more likely to search for brands they already know, according to a recent study by geoSurge. The analysis found that AI models chose familiar brands 3.2 times more often than unfamiliar ones when responding to U.S. buyer questions. In total, 55.7% of searches targeted brands within a model’s top 10, while only 17.4% focused on brands outside that group.
Researchers separated the concepts of model memory and search behavior to determine how prior knowledge shapes AI-driven discovery. Out of nearly 4,000 responses to 66 buyer prompts, the study revealed that 63% of brand-specific searches involved one of each model’s five most familiar brands. However, most fan-out searches-searches that explore multiple options-did not mention any brand at all, with only 31% including a company name.
Industry-specific results showed wide variation. Depending on the sector, AI models searched for familiar brands between 41% and 82% of the time, compared to just 9% to 23% for less familiar brands. The research covered industries such as travel, automotive, finance, business software, education, food and restaurants, luxury, fitness and wellness, and fashion. Some categories, however, were represented by as few as six prompts, which may affect the reliability of those findings.
Despite the strong influence of memory, the study found that AI models can still surface unfamiliar brands in certain cases. For example, Gemini identified Lemon Squeezy as an online payment provider, even though the brand was not part of its measured memory. Researchers noted that live search capabilities allow models to discover new brands, especially in sectors where prior knowledge is less dominant.
The implications are significant for brands seeking visibility in AI-powered search. The study suggests that established brands may enjoy an inherent advantage, as AI models are predisposed to search for names they already recognize. While strong content can help emerging brands gain attention, the data indicates that familiarity remains a powerful factor in AI-driven discovery. This echoes findings from other research, such as an analysis showing that most ChatGPT topic categories lack a clear brand leader, which can be explored in more detail in this related report on unsettled AI search topics.
geoSurge’s study was based on 66 U.S. buyer prompts, each tested 60 times between May 29 and June 9. In total, researchers analyzed 3,960 responses, 13,281 fan-out searches, and 1,416 brand-level observations. The authors emphasized that while there is a clear relationship between model memory and search behavior, the data does not prove causation.
