What to do on your phone instead of scrolling

Worldwide, the average mobile user spent 3.6 hours a day in apps in 2025.^[5] That is time inside iOS and Google Play apps, not everything a phone gets used for. You are not going to get that number to zero, which is the honest starting point most advice about screen time skips. So the useful question is not how to use your phone less. It is what to do with the four or five minutes of it you were going to spend anyway.

This piece used to be a list of opinions that sounded right. We have rewritten it around four papers and one industry estimate, and where the research did not support what we had written, the claim is gone rather than softened. There is a short list at the end of what we cut and why.

Why the replacement has to be small

This part is an argument rather than a finding, so read it as one. The reason scrolling wins is not that it is fun. It is that it costs nothing to start. There is no decision, no setup, and no point at which you have committed to anything. Anything that replaces it has to clear the same bar. A twenty-minute video course does not clear it. A book on the nightstand does not clear it, which is why it is still on the nightstand.

The replacement has to be startable in the four seconds you have while the lift arrives. Everything below is about what should happen in those four minutes once it has started, and that part is not a matter of opinion.

Answering beats re-reading, but not immediately

The single most replicated finding in this area is unhelpfully named. It is called the testing effect, which makes it sound like it is about exams. It is not. It is about the difference between putting information in front of your eyes again and making yourself produce it from memory.

In 2006 Roediger and Karpicke had students read a short prose passage, then either restudy it or take recall tests on it with no feedback, and then tested everyone again either five minutes later or a week later.^[1] Five minutes out, restudying won. A week later the ranking had reversed completely.

The group that read the passage four times and never tested themselves recalled 40% of it a week later. The group that read it once and then tried to recall it three times got 61%. They had spent less time looking at the material and remembered half again as much of it.^[1]

And it will not feel like it is working

This is the part that matters for anything you would actually use on a phone. Before the final test, the same students rated how well they expected to remember the passage in a week. The students who had restudied without testing rated themselves highest. They then remembered the least.

The authors put it plainly: repeated studying inflated students’ confidence in their ability to remember the passages, even though those students actually showed much poorer retention on the delayed tests.^[1] Re-reading feels like learning because the material gets easier to read each time. Fluency is not the same thing as memory, and it is the feeling most study habits are optimised for.

So a product that makes you answer questions will feel worse in the moment than one that just shows you a nicely written summary. That is a genuinely difficult thing to sell, and it is the correct design.

Come back to it later, not sooner

Retrieval is one half. The other is when the second pass happens. The structure of the question looks like this, and the finding concerns the first gap, not the second.

Cepeda and colleagues taught more than 1350 people a set of trivia facts, varied the gap before a review from nothing at all up to 105 days, and then tested them anywhere from a week to a year afterwards.^[2] Leaving a gap beat reviewing immediately at every single delay they tested, and the benefit grew with how long people needed to hold onto the material.

The practical shape of it: the best gap was roughly 20% of how long you need to remember something for delays of a few weeks, falling to around 5% at a one-year delay.^[2] If you want to still know something next month, the review wants to be days later, not minutes later. Which means a review queue has to exist, and has to be willing to interrupt you on a day you were not thinking about the topic.

It matters that you chose it

There is a real difference between what an algorithm hands you and what you went looking for. A 2008 meta-analysis of 41 studies found that giving people a choice enhanced intrinsic motivation, effort, task performance and perceived competence.^[3] We are citing that one carefully: we read its abstract but could not get hold of the full results, so we are not going to quote an effect size at you.

Both an algorithmic feed and a topic you typed in yourself can be interesting. Only one of them leaves you with the sense that you spent the time rather than lost it.

What does not work

Screen time limits, on their own. In a longitudinal field study of 242 people, published in the Journal of the Association for Consumer Research, screen time tracking improved how well people understood their own usage without reliably reducing it.^[4] Tracking tells you what not to do at the exact moment you have nothing else to do. The limit was never the missing piece.

Switching to a “better” feed. Swapping a doomscroll for an educational feed keeps the two parts that were the problem: it does not end, and you did not choose it. Nothing in the research above applies to material you passively watched once and never had to produce again.

Anything that requires a desk. If it does not survive a bus, it will not survive a Tuesday. That one is opinion, but we are fairly confident about it.

What we built, and what it does not do

NanoLearner is our attempt at the version of this that clears the four-second bar. You type in whatever you are curious about and get a lesson you can finish before your stop. There are quizzes, which is the retrieval half, and it comes back a few days later, which is the spacing half. The free tier gives you three topics, which is enough to find out whether this is a thing you will do.

It will not fix your relationship with your phone. Nothing installed on the phone does that. What it can do is make the ten minutes you were going to spend anyway add up to something by the end of the month.

Being clear about the limits: it is AI-generated, and we are more careful about that than most, but it is not a textbook and it is not a teacher. It is a good five minutes.

What we cut

The first version of this article claimed that almost nobody is still using screen time apps by March, and that most people tap straight through their own limits. Both sounded true. Neither had anything behind it, so both are gone. The second is now a narrower claim with a paper attached.^[4]

We also wanted to tell you how large the effect of choice is, and cut the number when we could not read the results section that contains it. Every figure above plots digits that are printed as numerals in the body of the cited paper, not values we estimated by eye off a published chart. Each figure names the page it came from, and you can open the numbers underneath it.