Every claim, with its receipt
ScreenLimit shows one short fact per pause. Each one is labelled with how strong the evidence behind it actually is, because “a randomised trial found” and “two things go together” are not the same claim.
- Strong · randomised trial · 10
- Study · association · 3
- Early · small study · 2
None of this is medical advice, and none of it is a statement about you personally. It is research about groups of people.
Sleep
- Strong · randomised trial
Students who kept their phone under 2 hours a day for 3 weeks slept 18% better and had 27% fewer depressive symptoms.
Source: BMC Medicine, 2025
- Early · small study
Students who put their phone away 30 min before bed fell asleep about 12 min faster.
Source: PLOS One, 2020
Mood
- Study · association
People who cut their phone use by about 2 hours a day walked around 800 more steps.
Source: PLOS One, 2024
- Strong · randomised trial
People who blocked mobile internet for 2 weeks felt better and focused better. An hour offline tonight?
Source: PNAS Nexus, 2025
- Strong · randomised trial
Cutting back on social media helped mental health about as much as quitting completely.
Source: Meta-analysis, 2025
Attention
- Strong · randomised trial
Two weeks without mobile internet improved attention as much as winding back 10 years of age-related decline.
Source: PNAS Nexus, 2025
Time
- Strong · randomised trial
When people cut their phone use, the gains came from more time with people, moving and being outside.
Source: PNAS Nexus, 2025
Scroll
- Strong · randomised trial
About a third of social media use comes from habit and self-control, not free choice.
Source: American Economic Review, 2022
- Strong · randomised trial
People who could set their own limits in advance used social media noticeably less.
Source: American Economic Review, 2022
- Strong · randomised trial
In research, people using a pause app opened their apps 57% less often after 6 weeks.
Source: PNAS, 2023
AI & thinking
- Strong · randomised trial
Students who practised maths with ChatGPT scored 17% worse on the exam once it was taken away.
Source: PNAS, 2025
- Strong · randomised trial
An AI tutor that gave hints instead of answers did not cause that drop. How you use AI matters.
Source: PNAS, 2025
- Study · association
Frequent AI use goes together with weaker critical thinking, through handing off the thinking.
Source: Societies, 2025
- Study · association
The more people trusted AI, the less critically they thought. Self-confidence worked the other way.
Source: Microsoft Research, 2025
- Early · small study
People who wrote essays with ChatGPT could barely quote their own work.
Source: MIT Media Lab, 2025
How we choose them
- A randomised trial beats a correlation, and we say which one it is.
- We do not round a finding up. If a study says 17%, the app says 17%, not “nearly a fifth”.
- We include findings that complicate the story — for example that an AI giving hints instead of answers did not hurt exam results. How you use a thing matters.
- No fact is shown twice within a week, so the app cannot nag with one number.