Jiang, Park, Xiao & Zhang (2025) · NBER Working Paper 33536 · February 2025
AI & the Extended Workday
Time-use evidence on AI exposure and working hours.
+3h
more weekly work in highly AI-exposed occupations relative to less-exposed ones after ChatGPT
American Time Use Survey diaries, 2022 vs 2023: moving from low to high generative-AI exposure across occupations is associated with about three more work hours and three fewer leisure hours per week.
Hourly wages rise with exposure, while job-satisfaction and work-life-balance ratings fall slightly.
The pattern is consistent with AI complementing labor: each hour becomes more productive, and firms demand more of them.
In a separate test of the monitoring channel, hours rise after 2020 for employees in remote-feasible jobs but not for the self-employed.
Why it matters
If AI makes each hour of work more valuable, longer workdays are the default outcome; whether workers share the gains depends on their bargaining position, not on the technology.
NBER working paper, not yet peer-reviewed; a relative comparison across occupations based on repeated cross-section time-use data. Source: Jiang, Park, Xiao & Zhang (2025), AI and the Extended Workday. NBER WP 33536.