AI Accurately Estimates Facial Aging From Photographs

AI FACIAL aging assessment accurately graded visible aging signs and predicted perceived age from standardized photographs.

Researchers developed an AI system capable of evaluating 13 facial aging signs and estimating visible age, potentially offering dermatologists a more standardized approach to assessing facial appearance. The findings suggest that automated image analysis could support reproducible outcome measurements in dermatology research.

AI Facial Aging Assessment Evaluated in 468 Women

The study analyzed standardized full-face photographs of 468 North American women aged 18–80 years recruited in Dallas and Phoenix.

Fifteen trained experts, including dermatologists, assessed 13 facial aging signs using established Skin Aging Atlas scales. Separately, 253 individuals aged 20–60 years estimated each participant’s visible age, defined as the age perceived by observers rather than chronological or biological age.

Researchers developed AI models using facial landmarks to isolate relevant facial regions. Deep learning algorithms were trained to reproduce averaged expert assessments, with 90% of the images used for model development and the remaining 10% reserved for evaluation.

AI Matches Expert Facial Aging Assessments

The AI system demonstrated strong agreement with expert evaluations, achieving mean absolute errors below 0.3 grading units for 11 of the 13 facial aging signs.

Across ancestry groups, approximately 98% of automated assessments for women of East Asian ancestry, 96% for women of African ancestry, and 97% for women of European ancestry fell within one grading unit of expert reference scores.

Human-assessed visible age correlated strongly with chronological age (r=0.94; P<.001), although participants were perceived as approximately 10 years older than their actual age on average.

AI-predicted visible age also closely matched human assessments (r=0.93; P<.001), with approximately 70% of predictions falling within five years of observer estimates.

Implications for AI in Dermatology

These findings highlight the potential of AI facial aging assessment to reduce observer variability and provide consistent measurements of visible facial changes.

For dermatology researchers, automated grading could facilitate longitudinal monitoring and standardized evaluation of facial aging outcomes without requiring repeated assessments by multiple experts.

However, the system was evaluated internally using standardized photographs of women. External validation across independent populations, men, and different imaging devices remains essential before clinical implementation.

Importantly, predicted visible age reflects measurable facial characteristics rather than biological aging or overall health. Further research incorporating biological biomarkers is needed to establish whether these measurements have broader clinical significance.

Reference
Flament F et al. AI-Based Grading of Facial Aging Signs and Prediction of Visible Age From Standardized Facial Images: Algorithm Development and Internal Evaluation Study. JMIR Dermatol. 2026;9:e93236. doi:10.2196/93236.