Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models

Research Question and Approach

This study investigates how large language models (LLMs) act as competitive gatekeepers by analyzing which brands they recommend in response to category-based queries. The author seeks to map the competitive structure of these recommendations across five industries—SaaS, consulting, fintech, e-commerce, and healthcare technology—using 3,750 responses generated by GPT-5.2, Google Gemini 3 Flash, and Perplexity sonar-pro. To ensure robustness, the study employs a "dice-roll" protocol, repeating each query five times per model to account for the stochastic nature of LLM outputs.

Key Metrics for AI Intelligence

The paper introduces three exploratory metrics to quantify brand ownership in AI outputs:

• Category Ownership Index (COI): Measures a brand's share of mentions within a specific category.
• Competitive Vacuum Index (CVI): Identifies categories where no single brand dominates.
• Displacement Score (DS): Quantifies whether the presence of one brand suppresses the mention of another (asymmetric substitution) or if they are recommended together (co-recommendation).

The results challenge the common narrative that AI recommendation markets are inherently winner-takes-all. The mean Gini coefficient across industries was 0.28, well below the 0.60 threshold typically associated with power-law distributions. Competitive vacuums were rare, appearing in only 8% of queries, indicating that models generally provide at least one recognized brand. However, cross-model agreement on the top-recommended brand was only 41.6%, suggesting that a brand's visibility is highly platform-dependent. Displacement patterns varied by industry: consulting favored co-recommendation (where major firms are often listed together), while other sectors like e-commerce and SaaS showed stronger one-directional displacement, where the presence of a leader often excludes challengers.