Overview

• A Nature study published October 8 analyzes nearly 1.4 million images and videos plus text to show systematic age–gender distortions across occupations.
• More than 6,000 human coders rated women depicted in images and YouTube videos as younger than men in comparable roles, with the gap largest in prestigious positions such as doctors and CEOs.
• In a randomized experiment, participants exposed to images of female workers estimated occupation-wide ages about two years younger, whereas exposure to male images nudged estimates slightly older.
• When generating and scoring about 17,300 resumes per gender across 54 jobs, ChatGPT made female candidates an average of 1.6 years younger, described them as less experienced, and assigned lower scores.
• The distortions do not match labor demographics, as U.S. Census data show men and women workers are similar in age, prompting calls for intersectional bias mitigation in AI and online ecosystems.