VERASOMA

Research

Advancing facial analysis and beauty science to lead the future of personalised, data-driven aesthetics.

Current study

How do large language models perceive facial attractiveness?

Our research team is investigating how popular LLMs (Claude, ChatGPT, Gemini, and Grok) assess facial attractiveness compared to humans, through a quantitative analysis of existing data. This project documents similarities and differences between AI and human perception of facial attractiveness, explores potential biases in how LLMs assess facial characteristics, and advances our understanding of potential use cases for AI in facial aesthetics.

Read the working notes Request the methodology PDF

How we work

  • 01

    Reproducible

    Every claim in a VERASOMA report ties back to a published, citable study or to a measurement we can re-run on demand.

  • 02

    Cohort-aware

    Comparisons run against demographic cohorts (age band, sex, ethnicity) rather than a one-size-fits-all average. The cohort is always disclosed in the report.

  • 03

    Open about limits

    Aesthetic perception is culturally and individually variable. Our analyses describe what the literature says — not what you should look like.