Caffeine is the most widely used psychoactive substance in the world and is known to disrupt healthy sleep. However, very few studies have directly tested the effect of caffeine abstinence on sleep, and these have yielded inconsistent findings. The purpose of the present study was to examine changes in sleep following caffeine abstinence and examine the extent to which characteristics of habitual caffeine use moderated this change. Participants included 66 healthy, young adults with habitual caffeine use and poor sleep. During the 2-week baseline, sleep was assessed using wrist actigraphy and daily caffeine use was assessed with bedtime diaries. Eligible participants then completed 1 week of caffeine abstinence, during which sleep was measured with wrist actigraphy. Multilevel models found no significant differences between either mean levels or growth trajectories of total sleep time or sleep efficiency between baseline and caffeine abstinence. Mean levels of sleep onset latency also did not differ between baseline and caffeine abstinence. A small but significant quadratic effect was observed, such that sleep onset latency decreased during the first few days of caffeine abstinence, then increased to levels above baseline. Characteristics of caffeine use did not moderate changes in sleep between baseline and caffeine abstinence. These data suggest that abstaining from caffeine may not result in long-term sleep improvement for habitual caffeine users, which contradicts the common sleep health recommendation. The present findings encourage more rigorous investigation of the effectiveness of caffeine restriction on sleep.