How to Check Your Phone Less: What the Evidence Actually Supports

The evidence retrieved for this article does not support any single tactic. In the retrieved studies, every change tested on its own against objectively logged checking, grayscale, notifications off, notification batching and usage pop-ups, left checking frequency unchanged. The two studies that did reduce checking bundled several changes at once and cannot say which one worked.

Four research teams have taken a single change to a smartphone, applied it to real people for days or weeks, and then counted the resulting unlocks with software instead of asking anyone to remember. Grayscale. Notifications switched off. Notifications rescheduled rather than eliminated. On-screen pop-ups reporting a person’s own usage back to them. Not one of those changes moved how often people picked up the phone (Dekker & Baumgartner 2024; Dekker, Baumgartner, Sumter & Ohme 2024; Fitz et al. 2019; Loid, Täht & Rozgonjuk 2020).

That is the result of the documented literature retrieval for this dossier, dated 4 September 2026, not of a systematic review, and it applies to the retrieved studies rather than to all research everywhere. Within that set, the score on isolated tactics measured against objectively logged checking is four attempts and four nulls. None of the pages retrieved on the first page of this search reports it, which is why it is the first thing here.

The number everyone quotes is counting something else

The most repeated figure about phone checking is 34 times a day, and it is a misreading of the paper it comes from. In that study, 34.11 is the median number of usage sessions logged. The same paper’s measure of habitual checking behaviour, the short unprompted inspection that most people mean by the word, averaged 3.39 a day (Oulasvirta, Rattenbury, Ma & Raita 2012). Two different quantities, one of which travels and one of which does not.

The durable contribution of that paper is mechanistic rather than numeric. Combining behavioural logging, a controlled field experiment and diaries (136 smartphone and 160 laptop users in the first study, 15 and 12 people in the two that followed), it described checking as a habit: a brief, repeated inspection triggered by cues and reinforced by an informational reward that arrives quickly. Moderate evidence, and the documented retrieval located no direct replication. The word that matters for everything below is habit: the behaviour is described as cued, not as decided each time.

Four tactics were isolated, logged, and none of them moved the count

Grayscale. A preregistered within-subjects study put 84 Dutch Android users, mean age 21.95, through six days of normal use followed by six days with the screen in grayscale, with unlocks logged by a research app. Unlocks were 80.03 a day (SD 30.75) at baseline and 80.33 a day (SD 30.07) in grayscale: t = 0.18, p = .860, d = 0.04. Total screen time did fall over the same period, and self-reported stress fell with it (Dekker & Baumgartner 2024). That contrast is the whole point. Grayscale shortened the sessions and left the reaching for the phone exactly where it was. One study, one country, Android only, a young sample, no located replication.

Notifications off. A preregistered randomised controlled trial of 205 people, one week long, with behaviour objectively logged alongside mobile diaries, found no effect of turning notifications off on checking frequency, screen time, perceived control, phone overuse, vigilance, productivity or distraction. Two things did shift: self-reported checking habit strength fell, and fear of missing out rose. The authors’ conclusion is that “these findings challenge the assumption that notifications play a prominent role in driving smartphone use and influencing user experiences” (Dekker, Baumgartner, Sumter & Ohme 2024). On the direction of the relationship they add that notifications “do not necessarily drive smartphone use, but they may reflect how extensively people use their smartphones.”

Notification timing. A four-arm randomised field trial ran for three weeks with objective Android logging: 237 participants of 333 recruited, in India, mean age 30.3, 19% female, and not preregistered. The omnibus test on unlocks was not statistically significant (p = .059). What did separate the arms was how people felt. Eliminating notifications entirely raised anxiety (d = +0.56) and phone-related fear of missing out (d = +0.59), with FoMO fully mediating the anxiety rise. Receiving the same notifications in three batches a day lowered stress (d = -0.56), improved mood (d = +0.49) and reduced inattention (d = -0.65) (Fitz et al. 2019).

Usage pop-ups. The fourth study isolated on-screen pop-ups reporting the user’s own usage back to them, with objective logging over two months, and it too did not reduce phone checking behaviour (Loid, Täht & Rozgonjuk 2020).

Two of the four are preregistered, one states that it is not, and the preregistration status of the fourth was not established here. All four logged behaviour rather than asking about it, which removes recall error as an explanation for the nulls. It does not establish that any of them was large enough to detect a small effect; adequate measurement and adequate statistical power are different things, and none of these studies is presented here as having ruled an effect out.

The two interventions that did move checking cannot tell you which part did it

One randomised within-subject cross-over trial of young adults in Singapore, using daily diaries alongside objective usage data, applied four changes at once: disabling Face or Touch ID, lengthening the passcode, switching the screen to grayscale, and moving social apps off the home screen. Checking frequency went down, and self-reported stress went up at the same time (Kasturiratna, Chua & Hartanto 2025). Those are directions, not magnitudes: the full text is paywalled and only the direction of effect was verified for this dossier, so no figure from that study is printed here. The authors attribute the stress rise to the frustration of partial restrictions without full disengagement. Because the four changes were applied together, the study cannot attribute the reduction to any one of them, and the one component that has been isolated, grayscale, left unlock counts unchanged on its own.

The other positive result did not measure checking at all. Olson and colleagues tested a ten-strategy package in a preregistered randomised trial against an active control, 70 of 82 participants analysed, in Montreal, iPhone only, mean age about 21 (Olson et al. 2022). It reduced time spent. It did not record pickups or checking frequency, and participants who did not meet the strategy-adherence criterion, following at least five of the ten strategies, were excluded from that analysis. Total use is a different outcome from checking, and this trial speaks only to the first.

Three of the four changes in that bundle have not been tested on their own

Disabling biometric unlock and lengthening a passcode are among the most confidently repeated pieces of advice on this subject. The documented literature retrieval for this dossier, dated 4 September 2026, located no study testing either tactic in isolation, on any outcome. They appear only inside bundles, and bundles cannot separate their components. The honest status of “turn off Face ID” is not that it has failed, but that no qualifying study testing it alone was located.

The most popular advice of all rests on a contested effect

Ward, Duke, Gneezy and Bos (2017) reported that the mere presence of a person’s own smartphone reduces available cognitive capacity, across two experiments with 520 and 275 participants, using a salience manipulation, not preregistered. The one preregistered direct replication located, with 511 participants, more than the original’s second experiment, did not reproduce the effect (Ruiz Pardo & Minda 2022). The meta-analytic picture is contested: one meta-analysis pooling 22 studies and 43 effects reports a very small effect, g = -0.14 (Böttger, Poschik & Zierer 2023), while another, covering 56 studies and 7,093 participants, finds an effect only for working memory and describes the literature as underpowered (Parry 2024). At best this is a very small effect that two syntheses disagree about, and it is worth being precise about what it is an effect on: this literature measures performance on cognitive tasks, not how often anybody checks a phone.

Friction has a documented cost

Across the retrieved studies, friction changed how people felt in both directions, and which direction depended on the design. Turning notifications off raised fear of missing out, eliminating them entirely raised anxiety, batching them lowered stress and improved mood, the four-change bundle raised stress while reducing checking, and grayscale lowered stress while leaving unlocks alone.

Friction can carry psychological costs and is not reliably cost-free. In the retrieved studies, the size and direction of the cost did not follow how drastic the change felt.

What the evidence does and does not support

Planning. The largest and most rigorous test located is a preregistered two-arm randomised trial (NCT04550286) run with German university students over 21 days, Android only, with smartphone use logged by app through Murmuras rather than self-reported. 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. 787 students were randomised, 389 to the intervention and 398 to the control. The objectively logged smartphone-usage analyses drew on the 555 who provided app-measured usage data, with individual models running on between 497 and 555 participants. The Time × Group interaction terms were not significant: b = 3.90, p = .44 for overall smartphone usage, and b = -1.65, p = .39 for social media usage (Brockmeier et al. 2025). That trial did not show that its action-planning and coping-planning intervention reduced total smartphone usage. It did not measure checking frequency at all, so it says nothing either way about the outcome this article is about.

The mechanism, not the slogan. A review of behaviour-change research presents changing the situation as a way to reduce how much a behaviour depends on moment-to-moment intention, and presents situational strategies as a complement to willpower rather than a replacement for it (Duckworth, Milkman & Laibson 2019). It is a review, not an experiment, and it is not about phone checking. The situational half of that argument, applied to the settings a phone arrives with, is handled on our page about why the defaults are doing the work. The documented retrieval located no head-to-head study comparing environmental change against intention for phone use, so any claim that one beats the other is not resting on evidence located here.

Stopping. No study located here shows cessation. The trials that reduced anything reduced it by a margin, from a high baseline, among people who stayed with the change.

So the useful split is between two goals usually discussed as one. Time spent and the unprompted glance are different outcomes, and in the retrieved studies they did not move together: in the grayscale trial one fell while the other did not budge. Nothing tested in isolation against logged behaviour moved the glance, and that is the finding to carry into anything that promises otherwise.

What remains is not nothing. Checking is described in the research as a cued habit, and cues are individual: a place, a time of day, an empty few seconds, one app on one screen. The evidence retrieved here cannot tell you which are yours, and generic tactics applied without knowing them have not moved the count.

If the next question is about tools that block things, that is a different question with a different evidence base, and it is handled on our page about focus apps that actually block distraction.

The wider evidence on attention and distraction, including what it does not settle, is set out in the digital distraction and focus guide.

HOS keeps checking this literature as it changes, including the results that do not support the advice already in circulation. The Weekly System has not launched yet. You can join the launch list.