Digital Detox: Cutting Down Works a Little, Cutting Off Does Not

The digital detox has a rare property among wellness interventions: it has been tested a great deal, including in one of the largest field experiments ever run on social media.

The results are consistent enough to state simply. Reducing your use produces a small, real improvement. Stopping entirely produces nothing - and occasionally produces the opposite.

That distinction is the whole finding, and it is almost exactly backwards from how detoxes are sold.

Two meta-analyses, published the same year, disagreeing usefully

Lemahieu and colleagues (2025) pooled 10 studies and 4,674 participants on social media abstinence. All three outcomes were null:

Outcome	Effect (g)	95% CI	p
Positive affect	0.03	-0.11 to 0.16	.69
Negative affect	-0.01	-0.13 to 0.10	.78
Life satisfaction	0.03	-0.17 to 0.22	.75

Duration of abstinence moderated nothing. Publication-bias testing was non-significant and trim-and-fill left the results unchanged. Their summary: abstaining “does not make you feel better; nor does it make you feel worse.”

Burnell and colleagues (2025) pooled 32 articles, 5,544 participants and 91 effect sizes - all randomised trials - on social media restriction. Here there was something:

• Overall wellbeing g = 0.17, 95% CI [0.08, 0.27]
• Positive indicators g = 0.17 [0.04, 0.29]
• Negative indicators g = 0.18 [0.08, 0.27]

Small, but real and reasonably precise.

The two syntheses are not contradicting each other. They are measuring different interventions: cutting down versus cutting off. And the one that works is the less dramatic one.

Burnell’s own recommendation makes the point sharper than we could: future work “should focus on these mechanisms, rather than broadly restricting time spent using social media.”

The largest experiment, and what it cost

Allcott, Braghieri, Eichmeyer and Gentzkow (2020), in the American Economic Review, paid people to deactivate Facebook for four weeks. The impact sample was 1,661 participants, with 90% verified compliance.

The wellbeing effects were positive and small:

Measure	Effect
Subjective wellbeing index	+0.09 SD
Life satisfaction	+0.12 SD
Happiness	+0.08 SD
Depression	+0.09 SD
Political polarization	-0.16 SD
Facebook use	-59.6 minutes/day
News knowledge	-0.19 SD

Three things in that table deserve more attention than they get.

The wellbeing gain was smaller than the polarization gain, and smaller still than the knowledge loss. Deactivating made people slightly happier, meaningfully less polarised, and meaningfully less informed. That is a trade, not a win.

The authors put the effect in context themselves: the wellbeing improvement was only 25-40% of a typical psychological intervention effect, and the cross-sectional correlation between Facebook use and wellbeing is about three times larger than the experimental effect. If you have read that heavy users are much unhappier, that gap is why - most of that correlation is not caused by the using.

The effect outlasted the experiment. A month later, participants were still using Facebook about 12 minutes a day less than controls, and 5% were still deactivated three months on.

One more result that undercuts a popular theory: Facebook turned out to be a complement to other social media, not a substitute. People did not migrate elsewhere. And there was no active-versus-passive-use heterogeneity, which is the dominant explanatory framework in the field.

The trials where it went the other way

This is the part that does not appear in detox marketing.

Vally and D’Souza (2019) randomised 68 participants to abstain from social media. Relative to controls, abstainers showed a decline in life satisfaction, an increase in negative affect, and an increase in loneliness. Their conclusion: abstinence “may not necessarily exact positive changes.”

Vanman, Baker and Tobin (2018) gave 138 participants five days without Facebook and measured salivary cortisol alongside self-report. The result splits against itself: physiological stress fell and life satisfaction fell with it. Coverage of that study reported the cortisol half.

Radtke and colleagues’ systematic review (2022) - 21 studies, 3,625 participants - found the same pattern across the field. Use went down in every study, so the manipulations worked. Depression symptoms declined in all three studies measuring them. But life satisfaction split four positive, two negative, two null; cognitive and physical performance were consistently null; and the review documents increased withdrawal symptoms, elevated negative affect, and heightened loneliness following detox. Only 6 of the 21 studies met their high-quality criteria.

The study that shows why the control group matters

Hunt and colleagues (2018) is the most-cited trial in this area, and the popular account of it drops the most interesting result. 143 undergraduates were randomised to limit Facebook, Instagram and Snapchat to ten minutes each per day, or to use them as usual, for three weeks.

The limited-use group showed significant reductions in loneliness and depression. That is the headline.

But both groups improved on anxiety and fear of missing out - which the authors attribute to self-monitoring, not to the restriction. Simply being asked to track your use did some of the work in both conditions.

That matters because a widely shared 2020 study reporting large detox benefits (fear of missing out d = 0.55, wellbeing d = 0.47, social connectedness d = 0.28, all p < 0.001, N = 61) used a within-subjects design with no control group. It cannot separate the intervention from time, expectancy and self-monitoring - precisely the confound Hunt’s control group exposed. It is a good illustration of the method; it is not evidence about detoxes.

What this does not establish

It does not say social media is harmless. Allcott found a real, if small, wellbeing effect and a considerably larger polarization effect.

It does not say the effects are the same for everyone. Tromholt’s study of 1,095 people found gains concentrated among heavy users, passive users, and those high in social-comparison envy. That heterogeneity is under-studied.

It does not cover phone use in general. Almost all of this literature is about social media platforms specifically, mostly Facebook, mostly for one to four weeks. Nobody has run a long-term trial.

Nothing here is clinical advice. If phone use is causing you real distress, that is worth raising with a professional - the point of this article is that the popular intervention is weaker than advertised, not that the distress is imaginary.

What the evidence actually supports

Cut down rather than cut off. This is the single defensible conclusion, and it comes from comparing two meta-analyses published in the same year: restriction g = 0.17, abstinence g ≈ 0.03 and non-significant.

Expect a small effect, and do not be disappointed by it. The best-run experiment produced 0.09 standard deviations - a quarter to two-fifths of a typical psychological intervention. The reason people expect more is that the correlation is three times larger than the causal effect, and the correlation is what gets reported.

Notice what you lose. The same experiment that made people slightly happier made them measurably less informed. A detox is not free.

Be wary of the withdrawal. Several trials found abstainers lonelier and more anxious than controls. If you have tried a detox and felt worse, that is a documented outcome, not a personal failure.

The most useful framing may be Burnell’s: the thing worth investigating is not how many minutes you spend, but what those minutes are doing. Time is the variable that is easy to measure, which is why it is the one everybody measures. It does not follow that it is the one that matters.

Digital detoxes sit alongside the other interventions people try - see the screens, attention and focus guide for the wider picture, how to use your phone less for what the intervention evidence supports, and how to check your phone less for the narrower behavioural version.