Background: Walking is known to be good for health. However, it is unknown whether it matters if steps are taken during work or leisure. Therefore, we aimed to examine the prospective association between accelerometer-measured steps taken during work or leisure and register-based long-term sickness absence (LTSA). Methods: We included 937 blue- and white-collar workers from the PODESA cohort who wore a thigh-based accelerometer over four days to measure number of steps during work and leisure. Steps were divided into domain based on diary data. First event of LTSA was retrieved from a national register with four years’ follow-up. We used Cox proportional hazard models to analyze the association between domain-specific and total daily steps and LTSA, adjusted for age, sex, job type, smoking, and steps in the other domain (e.g., work/leisure). Results: We found more steps at work to be associated with a higher LTSA risk [Hazard Ratio (HR):1.04; 95% CI: 1.00–1.08 per 1000 steps]. No significant association was found between steps during leisure and LTSA (HR: 0.97; 95% CI: 0.91–1.02), nor between total daily steps and LTSA (HR: 1.01; CI 95% 0.99–1.04). Conclusions: More steps at work were associated with higher risk of LTSA, while steps during leisure was not clearly associated with LTSA risk. These findings partly support ‘the physical activity paradox’ stating that the association between physical activity and health depends on the domain.