How do large language models distribute their attention across different occupational categories? At Fameproof, we audited 50+ global public figures across actors, musicians, athletes, and business leaders using standardized prompt templates. The results reveal clear clusters of occupational bias.
Our data shows that models do not treat all fields of public life equally. Some domains exhibit highly centralized visibility, while others are distributed and dynamic.
Centralization: The Winner-Take-All effect in sports
Athletes and sports figures exhibit the highest concentration of visibility scores. In football/soccer, for example, the top two figures (Cristiano Ronaldo and Lionel Messi) command a near-monopoly on general recommendations and general-recognition prompts. Their visibility scores consistently hover above 90/100, while the next tier of athletes drops off sharply to below 40/100.
This indicates that when models are steered towards sports, their probability weights are highly consolidated around a few landmark entities, leaving a very thin tail for other contemporary players.
Decentralization: The flat distribution of acting & cinema
In contrast, actors and filmmakers display a much flatter, more decentralized visibility curve. While household names like Tom Hanks, Leonardo DiCaprio, and Scarlett Johansson score highly, they do not dominate their category the way top athletes do. A general acting query returns a highly diverse selection of 10-name lists, with lower individual mention rates but higher collective variety.
This suggests that models have a much broader and more balanced index of notable actors in their pre-training data, leading to less positional concentration in generated lists.
Occupational steering vs. Biography keywords
Our research also highlighted the importance of 'occupational steering'. When a subject's description combines multiple roles (e.g., 'actor, singer, and entrepreneur'), models frequently route them into a single primary category. A subject categorized primarily as a businessman might receive zero mentions in actor prompts, despite having a massive filmography.
To optimize visibility, brands and public figures must maintain a clear, consistent primary context that matches the specific category domains models query for.