Why the ruler you use can change the answer
Self-reports ask people how much they used social media; device data records what the phone actually did. These are not the same quantity. A review of 14 studies on reducing social media use and sleep quality notes that most were largely cross-sectional and relied on self-reports of use, and that measured app use in the studies it examined averaged only about 16 minutes a day, which the authors read as a sign of limited compliance with the reductions [2]. In other words, a self-report can say "I cut back a lot" while the device shows only a small drop, and that gap matters when you try to estimate an effect.
The strongest design for separating measurement from effect is to record baseline use objectively and then randomize people to suspend social media or carry on as usual. One experiment did exactly that with 93 participants, capturing a week of smartphone social media use before randomly assigning half to suspend use for a week, then re-measuring anxiety and sleep quality with questionnaires [1]. It found no evidence of a causal effect of social media use on anxiety or sleep quality [1]. Because the exposure was measured by the device rather than recalled, this null result is harder to dismiss as people misremembering how much they used.
Device data mostly confirms that reductions are smaller than people think
The practical consequence of device measurement is that the "dose" of reduction turns out to be modest. The review of adolescent and college-student studies concluded that reducing social media use improves sleep quality, but its own compliance figures suggest the actual cut in app use was around 16 minutes per day [2]. That is a small dose, so any sleep benefit estimated from these studies describes the effect of a modest reduction, not of quitting.
This is where measurement choice changes the estimate rather than the underlying biology. If participants self-report a large reduction but the device shows a small one, a study relying on self-report will attribute the sleep change to a bigger cut than actually occurred, inflating the apparent effect per minute of social media removed. The experiment that used objective smartphone data and found no anxiety or sleep effect is consistent with the idea that when you measure the exposure precisely, the effect is weaker or absent [1]. The authors themselves raise the possibility that this reflects a genuine absence of a relationship rather than a failure to detect one [1].
Nighttime scrolling, sleep procrastination, and the anxiety loop
Anxiety enters this picture less as a direct consequence of scrolling and more as part of a loop around bedtime. A study of young adults using ecological momentary assessment, which prompts people repeatedly in daily life rather than asking them to recall later, examined the bidirectional relationship between nighttime social media use and sleep outcomes [3]. It frames fear of missing out, the anxiety that others might be having rewarding experiences without you, as part of this nighttime pattern [3]. That is a proposed loop rather than a demonstrated cause of anxiety symptoms.
The measurement lesson carries over. That study validated self-reports against device records in a randomly selected subset of 20 participants and tracked compliance as the proportion of scheduled prompts answered [3]. Repeated in-the-moment sampling reduces recall error compared with a single retrospective questionnaire, but it still depends on people responding, and the validation covered only a small subset. So the nighttime evidence is best read as describing a plausible cycle linking late social media use, delayed sleep, and worry about missing out, not as proof that cutting social media reduces anxiety.
What this means if you are judging the advice for yourself
The honest summary is that the measurement method changes the size and confidence of the estimated effect more than its direction. With objective device data and random assignment, a one-week suspension produced no detectable change in anxiety or sleep quality [1]. With mostly self-reported use, reviews report sleep improvements, but the measured reductions were small and compliance was limited [2]. Neither result supports a strong claim that cutting social media reliably improves anxiety.
Two gaps keep the question open. The experiment captured smartphone social media data only, and its authors note that future work should include computers and other devices, so some use may have gone unmeasured [1]. And the nighttime study's momentary design captures a bidirectional pattern in young adults rather than a tested intervention [3]. If you want to know whether reducing social media would help your own sleep or anxiety, the evidence is most consistent with expecting a modest, uncertain benefit at best, and with tracking your actual use rather than relying on your impression of it.
Sources used in this answer
Assessing the effects of acute reductions in mobile device social media use on anxiety and sleep
In a randomized one-week experiment with objectively captured smartphone social media use, suspending social media produced no evidence of a causal effect on anxiety or sleep quality.
To sleep–Perchance to dream: does reducing social media use improve sleep quality among adolescents and college students?
A review of 14 studies reports that reducing social media use improves sleep quality, while noting most studies relied on self-reported use and that measured app use averaged about 16 minutes per day, suggesting limited compliance.
The vicious cycle of nighttime social media use and sleep procrastination in young adults: an ecological momentary assessment study
An ecological momentary assessment study of young adults describes a bidirectional relationship between nighttime social media use and sleep outcomes, framing fear of missing out as part of the pattern, and validated self-reports against device records in a subset of 20 participants.