NameRank Measures LLM Recognition of People and Projects

A new study introduces NameRank, a quantitative score from 0 to 1, designed to assess what frontier large language models (LLMs) recall about individuals or tools from their internal knowledge, prior to any retrieval processes. The research probes 4,685 entities across 54 cohorts using open-ended questions with 36 different models. An independent judge then determines if the model provided a specific, non-guessable fact about the entity, filtering out hallucinations or context echoes. The key finding is that LLM recognition is predominantly tied to named, indexable artifacts rather than personal credentials or titles. For instance, while Nobel laureates are recognized, most other academic credentials fall below a baseline for working researchers because no specific named artifact is associated with the credential itself. Conversely, independent creators often see their tools recognized more than their own names. The study also reveals that being one of many contributors to a celebrated artifact, like authors on a flagship model report, yields minimal recognition; the recognition attaches to the artifact's distinct name. Traditional bibliometrics do not predict recognition well, and institutional prestige only helps at the very top tier. Recognition for news events peaks at the moment of salience rather than persisting over time.