In June 2026, the Medical Journal of Australia published a longitudinal study of 1,195 participants from Melbourne, recruited through schools and monitored from childhood until the age of nineteen. Between the ages of twelve and eighteen, they reported their daily social media usage each year, and their mental health was assessed in the subsequent annual wave. The data extends to 2022, prior to Australia prohibiting social media accounts for those under sixteen.
In the primary comparison, those who reported more than two hours of daily use, as opposed to those who reported less than one hour, exhibited a slightly higher estimated risk of high depressive symptoms and low well-being in the following year. The absolute difference was 6.3 cases per hundred for depressive symptoms and 4.9 for low well-being.
These are not adolescents for whom social media has caused a problem. These are estimated differences between two usage groups, following the application of the study's statistical model. Furthermore, these findings do not establish two hours as a clinical threshold: the research does not demonstrate that an individual transitions from a protected state to one of risk upon crossing that point.
The associations were more pronounced in early adolescence. The highest estimate was for depressive symptoms in girls aged twelve and thirteen who reported more than two hours of use: 10.8 additional cases per hundred compared to the group reporting less than one hour. The confidence interval, ranging from 2.7 to 18.9, serves as a reminder that the precise value remains uncertain and that analyses stratified by age and sex involve smaller cohorts.
I have dedicated many years to digital communication and the technology sector. I interpret this study through an understanding of platforms, rather than from clinical practice or mental health research.
What the study measures
It measures associations over time, not a proven cause. Exposure is the daily time each adolescent reported spending on social media, not a record taken from their devices. Nor does it distinguish between platforms, activities or contexts: an hour chatting with friends, an hour posting original content and an hour of recommended videos all count the same.
The results are also derived from questionnaires. Depressive symptoms, anxiety, well-being, and self-harm were evaluated based on participant responses. Across the adolescent period as a whole, the associations highlighted by the article are those concerning high depressive symptoms and low well-being, and these should not be extrapolated to the other outcomes studied without further evidence.
The models were adjusted for variables including prior mental health and social media usage, peer bullying, parental support, sleep, physical activity, age, sex, country of birth, and socioeconomic status. Additionally, missing data were addressed through multiple imputation: 945 participants, or 79.1%, lacked data for at least one variable across the eleven waves.
While this statistical rigour improves comparison, it does not transform an observational cohort into an experiment. Unobserved or incompletely measured differences may persist. Moreover, the relationship may function in both directions: the study sequences exposure and outcome in consecutive years and adjusts for prior measures, yet it does not entirely exclude the possibility that distress may lead to increased social media usage.
Stated precisely: under the assumptions of the model, reporting more than two hours of daily use was associated with small subsequent increases in the estimated risk of two specific outcomes, compared to reporting less than one hour.
Before the legislation
The data concluded in 2022, prior to the Australian restrictions that entered into force on 10 December 2025. Since that date, the included platforms may not permit individuals under sixteen years of age to create or maintain accounts, and they must adopt reasonable measures to prevent it.
The regulation does not necessarily prevent the viewing of public content without logging in. Furthermore, the study does not evaluate this legislation: it cannot determine whether the restriction will reduce the observed risks, whether it will shift a portion of usage to services not included, or whether it will have varying effects depending on age and context.
The Murdoch Children’s Research Institute itself is participating in further research tracking adolescents both before and after the implementation of the restriction. That study will be able to provide evidence regarding its effects. The 2026 study only describes what occurred beforehand.
Three questions to keep apart
Public discourse tends to conflate three questions:
- At what age should an account be permitted. This is the question addressed by the Australian regulation.
- How much time is spent on a platform. This is what the study measures.
- Which system characteristics produce or intensify specific usage patterns. This is a matter of design that the study does not observe.
These are related, yet one does not account for the others. An age restriction may reduce account-based access without addressing any mechanism of recommendation, notification, or reward. Furthermore, recommending less time does not specify which usage should be reduced or which portion proves harmful.
The authors consider that the results support policies that mitigate potential adverse effects, with particular attention to early adolescence. The Emerging Minds research summary mentions education, guidance for families, and platform safeguards. However, the study does not compare these interventions, and therefore does not allow for them to be ranked by efficacy.
Hours and design
In my view, hours dominate the debate because they are easily measured. They can be counted, compared, and converted into recommendations. Design is more difficult to isolate: it involves content sequencing, autoplay, notifications, infinite scroll, public metrics, personalisation, and commercial incentives. This study does not allow for an effect to be attributed to any of those elements.
This does not render time irrelevant. If an association appears when comparing durations, reducing exposure is a reasonable hypothesis for intervention. However, immediately converting it into individual responsibility leaves unexamined how that exposure is manufactured. The two questions should advance together: how much time is spent and what the system does to achieve it.
This is the argument I maintained in “Mass technology and ethical consumption”: within infrastructures of this scale, the capacity of an individual or a family to decide exists, but it operates within options organised by design, the business model, and regulation.
This does not absolve families of responsibility nor does it diminish a result that merits attention, although it must be noted that a law regarding age of access, a recommendation on time, and a standard regarding design are distinct interventions. The study provides limited evidence on the second and context to discuss the first. Regarding the third, it leaves the decisive question open.
What remains to be measured
To approach the design, data would be required that separates, at the very least:
- The platform and the account type.
- The activity performed.
- The content selected versus the content recommended.
- Active usage and passive consumption.
- Interaction with known and unknown individuals.
- Schedule, interruptions, and lost sleep.
- Notifications and the sessions initiated.
- Positive experiences, harassment, and harmful content.
- Device usage logs, alongside participant self-reporting.
No single variable would resolve the problem in isolation. However, they would allow us to cease treating forms of usage that are likely distinct as if they were identical.
About the open conversation
This text continues the Public Notebook series on digital infrastructures and structural responsibility. It addresses research on adolescent mental health without intending to provide clinical advice. If this subject affects you or you are experiencing it closely, speaking with a health professional may assist.
If you work in developmental psychology, public health, epidemiology or platform design and can correct a figure or discuss this reading, this notebook remains open: write to me.
Sources
Australian Government, Office for Youth, “Social media age restrictions”.