How to Use Your Phone Less: What the Evidence Supports

No single method has been shown to reduce total phone use reliably and durably. In the retrieved studies, some structural restrictions produced short-term changes or benefits, but adherence was poor and persistence was largely unmeasured. That is the honest state of the evidence, and the rest of this article shows how it was reached.

First, decide which of four things you actually want

Most advice answers all four at once. The evidence does not.

Total screen time is minutes on the device, and it is what most trials measure. Attention or well-being is a different outcome, measured differently, which does not automatically follow from fewer minutes. Checking frequency is the unprompted glance, counted in pickups and unlocks; it moves independently of screen time and is a separate question with its own evidence, covered in a forthcoming HOS article. Lasting change is whether anything survives the end of the experiment, and it has the least evidence behind it.

What the retrieved evidence says, by goal

This table reports what studies found. It does not name a winner, because the studies measured different outcomes and cannot be ranked against each other.

One thing the table deliberately excludes

Most guides on this subject also import social media abstinence research: quitting one platform for a fixed period. Quitting Facebook for four weeks says nothing about whether a screen-time cap changes total device use, because the behaviour, the cue, the substitution options and the participants all differ. None of it is used above, and importing it is the most common error in the advice coverage of this subject. It is handled separately at the end.

What the studies found

Everything in this section reports what a retrieved study measured. No recommendation appears here.

Planning did not move logged minutes in the largest trial retrieved. Brockmeier et al. (2025) randomised 787 German university students, 389 to the intervention and 398 to control. 716 installed the study app and 555 provided app-measured usage data, with individual models running on between 497 and 555 participants, analysed as intention-to-treat. The intervention was action planning, three plans specifying when, where and how long the phone would be put away, plus coping planning, three if-then plans anticipating barriers. Across 21 days of logged use the Time by Group interaction was b = 3.90, p = .44 for overall usage and b = -1.65, p = .39 for social media usage. Logging was Android only.

Blocking mobile internet produced improvements and poor compliance in the same study. In Castelo et al. (2025), a preregistered crossover trial, participants installed a blocker for two weeks. Sustained attention, mental health and wellbeing all improved, with within-person effect sizes of 0.23, 0.56 and 0.45. Only 25.5% of those who committed met the compliance threshold, and the results are intention-to-treat, so they include the majority who did not comply. Read literally, this is a finding about being asked to block mobile internet, not about blocking it.

A two-hour daily cap improved symptoms, in a design that cannot fully attribute the improvement. Pieh et al. (2025) randomised 111 Austrian students aged 18 to 29 into a three-week instruction to cut phone use to two hours a day. Depressive symptoms, stress, sleep and wellbeing all improved. Adherence came from weekly screenshots of the built-in tracker, the study was not blinded, and the sample was students. Those limits reduce how much causal weight the result carries. They do not show the instruction did nothing.

Grayscale produced short-term reductions across three studies that are hard to compare. Screen time fell in a quasi-experiment, a preregistered three-arm randomised field experiment and a device-logged within-subjects study, over windows of days to three weeks (Holte and Ferraro 2023; Zimmermann and Sobolev 2023; Dekker and Baumgartner 2024). The designs and measurement methods differ enough that no pooled estimate exists. It has a forthcoming HOS article of its own.

Timers and blockers were installed and then routed around. In a three-week Android field study of 38 people, 97.81% of 91 timer triggers and 100% of blocker triggers were snoozed, and every blocking app was deleted before the study ended. Unlocks moved from 121.84 to 116.64 a day, which is to say they did not move (Roffarello and De Russis 2019). The article on blocking tools covers what distinguishes a blocker people keep from one they delete.

A control group fell alongside the treatment arms. In a three-arm field experiment with 97 students over one week, comparing moving problematic apps to another home screen page, grayscale, and a control condition, use fell during the intervention and also fell in the control group, while self-reported problematic use rose (Ochs and Sauer 2022). That result is about that study, and it demonstrates that observation or self-monitoring can contribute to an early decline.

The detox literature as a whole is weak, and its own reviewers say so. A systematic review of 21 studies covering 3,625 participants found that 6 of the 21 met its high-quality criteria. It is a narrative review rather than a meta-analysis and produces no pooled effect size. Cognitive and physical performance outcomes were consistently null (Radtke et al. 2022).

The pooled estimates that exist are not estimates of daily minutes. Balhara et al. (2026) included 125 studies, 73 of them randomised trials, and reports pooled effects for smartphone addiction of -1.49 for psychological interventions and -3.07 for exercise-based ones. The authors report high heterogeneity and evidence of small-study effects, which is their own warning about the weight those figures carry, and state there were too few studies to pool a result for social media addiction. These are estimates for problematic-use constructs, not for daily minutes on a device.

Social media abstinence, kept separate

Allcott et al. (2020) paid 1,661 US participants to deactivate Facebook in late 2018. About four weeks after the paid period ended, use remained reduced by roughly 12 minutes a day, about 23%. Among iPhone users reporting directly from their Settings screen, the cleanest measurement in the study, the reduction was 8% and not statistically significant, and 95% were back within nine weeks. This is the only persistence measurement retrieved for this article, and it is about one platform rather than total phone use. The wellbeing question is unsettled even here: Tromholt (2016) found an improvement after a week off Facebook, and Hall et al. (2019) found no main effect.

Adherence, and what has not been measured

Adherence is the least-reported number in this literature and the one that decides everything. In the retrieved studies, roughly a quarter of committed participants met a blocking threshold, almost every timer and blocker prompt was snoozed, and every blocking app was deleted before its study ended. Those were volunteers who knew they were being observed.

That changes how to read any headline result. A trial reports what happened to everyone assigned to the intervention, including the people who did not do it. Intention-to-treat is the right way to run a trial and a poor way to predict what happens to one determined person. It cuts both ways: these studies also cannot say what happens under full compliance, because too few people complied.

Persistence is the larger gap. The documented retrieval for this dossier, dated 4 September 2026, did not locate a study of post-intervention persistence for total phone use: none finding that effects fade, none finding that they last. That is a statement about what was found, not proof that nothing exists. Either way, an article promising lasting change is asserting something the retrieved research has not measured.

How HOS reads this

Everything below this line is editorial interpretation. It is not a validated clinical protocol, and no study in this article tested it.

Two readings seem defensible. Techniques that depend on responding correctly to a prompt in the moment have the worst measured record in the retrieved studies. And compliance, rather than effect size, looks like the thing to plan around, because the trial with the clearest improvements also had the worst adherence.

If you want to act on that, treat it as an HOS evidence-informed self-experiment rather than a proven treatment:

• Define one measurable outcome, and only one.
• Record a baseline before changing anything.
• Test one reversible change for a fixed period.
• Measure adherence alongside the outcome, because adherence is the variable the research keeps

losing.

• Stop or modify the experiment if it produces meaningful stress or interferes with access you

actually need.

Nothing in the retrieved literature establishes that this will work or that any effect will outlast the period you set, so do not budget for permanence. Cheap, reversible and honestly labelled is the most this evidence supports.

The wider evidence on attention and distraction, including the parts this page does not cover, is set out in the digital distraction and focus guide.

HOS reads the studies and reports what they actually measured, including when the answer is inconvenient. The Weekly System has not launched yet. If you want it when it does, join the launch list.