From Prompt to Recommendation: A Fitted Stage Model of Brand Visibility in AI Search
Topics: Brand Context, LLMO / GEO, Retrieval Augmented Generation (RAG)
This research paper investigates what actually determines whether a brand appears in an AI-generated answer. Rather than assuming that having a well-optimized page is enough, the paper breaks down brand visibility into a series of distinct stages: whether the AI searches at all, whether it finds supporting evidence for the brand, and whether it ultimately mentions the brand in its answer. The study analyzes nearly 35,000 observations from real AI interactions across GPT (ChatGPT) and Gemini, and builds a predictive model that combines a brand’s historical appearance rate with live retrieval signals.
This paper applies directly to AI search systems including ChatGPT (GPT-4/5 variants) and Google Gemini, both of which are tested extensively throughout the research. It is highly relevant to GEO (Generative Engine Optimization) practitioners because it challenges the assumption that content quality alone drives AI visibility. Instead, it shows that where a brand appears in the AI’s retrieval process — not just how good the content is — is the dominant factor in whether a brand gets recommended.
