Category Ownership Map (COI/CVI/DS): A Multi-Industry Dataset of Brand Recommendations Across Large Language Models (2026)

Data and analysis code behind the study "Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models." The dataset contains 3,750 responses from three large language models (GPT-5.2, Google Gemini 3 Flash, Perplexity sonar-pro) to 250 brand-free category buying questions across 50 brands and five industries (SaaS, consulting, fintech, e-commerce, healthcare technology), with each query repeated five times under a dice-roll stability protocol. It includes the raw responses, extracted brand mentions, the Category Ownership Index (COI), Competitive Vacuum Index (CVI), and Displacement Score (DS) matrices, response embeddings used for BERTopic, the result tables and figures, and the analysis notebook that reproduces them. Headline results: mean Gini 0.284 (moderate concentration), 8.0% competitive vacuums, 4.2% emergent clusters, 41.6% cross-model agreement on the top brand.