Australia's Teen SNS Ban: Fewer Accounts, But Then What?

Fewer accounts, but the mental health effect needs separate verification

On December 10, 2025, Australia began requiring designated social media platforms to take reasonable steps to prevent users under 16 from creating or holding accounts. The following January, the government announced that roughly 4.7 million accounts had been deleted or restricted in the early stages of enforcement. That doesn’t mean 4.7 million children stopped using these platforms — a single person can hold multiple accounts.

A subsequent parental survey found that while account ownership had declined, a substantial share of existing users had kept their accounts. These aren’t contradictory findings. You just have to look at what population each figure is measuring against to make sense of it. The bigger question comes next: did the drop in account numbers actually translate into better mental health for kids?

4.7 Million and 70% Don’t Share the Same Denominator

4.7 million is the number of deleted or access-restricted accounts that platforms reported to the government through mid-December 2025. That figure can include accounts that were already inactive. It’s evidence that platforms took action, but it doesn’t tell us how actual user numbers or usage time have changed.

Australia’s eSafety Commissioner (eSafety) surveyed 898 parents and guardians of children aged 8 to 15 from January 19 to February 2, 2026. Among children who had accounts on these platforms before the regulation took effect, 69.1% for Instagram, 69.4% for Snapchat, and 69.3% for TikTok said their children still had accounts afterward. This does not mean roughly 70% of all teenagers have accounts.

In the same survey, the share of respondents saying their child had an account on at least one of the target platforms fell from 49.7% before enforcement to 31.3% after. This comes with the limitation of reflecting only what parents were aware of, but it can’t be read as “account ownership barely changed.” eSafety, March 2026 compliance update

These two figures need to be read together. Even if platforms restricted a large number of accounts, some children may still have accounts — either because existing accounts haven’t yet been identified, or because new ones were created. Account ownership, logged-out browsing, and total usage time are all different metrics.

What I’m wary of is treating enforcement counts as the final measure of success. The number of restricted accounts is useful for tracking implementation. But if that alone is used to claim children have become safer, it risks becoming a vanity metric¹ — a good-looking number that overstates real impact.

It would also be wrong to conclude that the government is hiding unfavorable numbers. It was eSafety itself that disclosed the roughly-70% retention rate. The agency has explained the gap between account counts and user counts, and it has announced a two-year tracking study to evaluate both the implementation and the impact of the policy. Now we need to watch how well that evaluation actually captures changes in real safety and well-being.

In Australia, courts can impose civil penalties of up to A$49.5 million (~$32 million) on platforms that fail to meet their obligations. This isn’t a mechanism for punishing children or parents. Given how strong the penalties are, the responsibility to assess what real difference the platforms’ actions are making is correspondingly greater.

Short-Term Adult Experiments Can’t Predict Teen Policy Outcomes

There isn’t enough evidence yet on how account restrictions will affect teenagers’ mental health. That doesn’t mean the conclusion is “no effect,” either. We need to look closely at what the existing research actually tested.

A team led by Monika Nef Lind at UC Irvine reviewed the existing experimental literature in a May 2026 paper. They examined 40 randomized controlled trials (RCTs)² that split participants into groups who cut back or quit social media versus comparison groups. They combined 36 studies from two prior meta-analyses with 4 additional studies found through supplementary searches.

When applying this to policy, you need to separate three things: who the participants were, how long the experiments ran, and how large the effects were.

None of the studies reviewed had an average participant age under 18, and the youngest reported participant was 16. These 40 studies cannot directly tell us how people under 16 — the group the regulation actually targets — would respond. This doesn’t mean there’s no teen research on social media at all; it means the usage-restriction experiments included in this review simply studied a different population.

Eight studies included some participants under 18. According to the research team, even these participants were recruited through university subject pools — which is why they’re hard to treat as representative of teenagers as a whole.

The experiments were also short. Restriction periods ranged from a single day to three months, averaging 16.3 days, with half lasting a week or less. The studies mixed complete abstinence with reduced usage. You can’t equate a brief, voluntary experiment with a multi-year, nationwide account restriction.

The average effects in the existing meta-analyses were either statistically indistinguishable from zero or small. One analysis reported an effect size³ of about 0.17. Among the 40 studies, more reported improvement than not, but some showed no difference or even worsening on certain measures. Because participants and outcome measures varied study to study, you can’t simply count studies and convert that into a policy “success rate” or “failure rate.”

Participants generally knew they were in an experiment designed to restrict their usage. The research team notes this expectation could have influenced how they responded⁴. But exactly how much this affected any given study hasn’t been established. And the finding that older-age samples showed larger improvement effects in the two meta-analyses is not, by itself, direct evidence that the intervention has no effect on people under 16.

The U.S. National Academies’ 2024 report likewise concluded that the relationship between social media and adolescent health resists simple conclusions. Depending on how it’s used, what content is consumed, and individual circumstances, harmful and beneficial experiences can coexist. A small average association doesn’t mean the harm experienced by any particular child is necessarily small.

There are also limits to how this is measured. When usage and depressive symptoms are asked about in the same survey, a correlation shows up but it’s hard to tell which came first. Factors like family environment — which can influence both — may not have been adequately accounted for. This is why you need to track changes over time alongside a comparison group.

The research team’s core argument is that any policy rollout should be paired with rigorous evaluation from day one. In other words: acknowledge the limitations of the existing experiments, but track the actual population targeted by the policy — teenagers — over the long term.

Some studies do show restriction helping. Davis and Goldfield’s study found that reducing social media use among 17-to-25-year-olds experiencing emotional difficulties lowered depression, anxiety, and fear of missing out. Results like this demonstrate benefit under specific conditions. But they don’t substitute for proof that a policy banning account ownership for everyone under 16 would work.

The draft looks accurate and complete. No Hangul, all numbers preserved, structure matches, link intact. No edits needed.

We should also watch how workarounds and protections shift

Evaluating a policy means checking side effects alongside the intended ones. What follows isn’t a set of outcomes confirmed to have already happened — it’s a list of risks worth monitoring once the law takes effect.

Australia’s obligation applies to account creation and holding on the platforms in question. It doesn’t block content that can be viewed without logging in. If a child watches while logged out, or uses an account registered as an adult, they may end up outside the reach of the content filters and parental controls built for teen accounts.

Age verification can rely on several methods — ID checks, face-based age estimation, or existing account data. The government requires that ID submission not be mandatory and that reasonable alternatives be offered. Even so, we need to check how much personal data gets collected and retained, and how errors in age assessment get corrected. Since accuracy varies by technology, we also need real operational data to see whether error rates run higher for particular groups. eSafety’s age-verification guidance

How teenagers respond to the rules matters too. Researchers worry that restrictions perceived as ignoring autonomy can trigger backlash or lead kids to hide their usage. We need to ask both children and parents whether that’s actually happening.

If schools or youth organizations use social media to send out notices and event information, alternative channels need to exist. Restricting accounts shouldn’t mean kids miss information they actually need.

There’s also a chance some teens migrate to other, less strictly moderated services. Looking only at declining usage on major platforms could miss this shift entirely. To know whether safety has actually improved, we need to track what content kids get exposed to and who they turn to for help.

Even if account numbers drop, usage patterns could shift in several directions at once. Logged-out viewing, new accounts, and migration to other services all need to be examined separately.

Oswarld’s Lens

There’s a failure pattern I’ve seen often when building GTM strategies: launching a solution before fully defining the problem, then announcing the launch itself as the achievement. With this policy too, we need to separate “how many accounts were restricted” from “did children’s safety and mental health actually improve.”

There’s a question I always ask when working with data: “What exactly does this number measure?” In the Australian case, we need to read separately the number of deleted or restricted accounts, the retention rate of existing users’ accounts, and the account-holding rate across the entire surveyed population. Because you can package the same policy as a success or a failure depending on which number you pick, I always start by checking the denominator behind a published figure and what measurement is missing from it.

The paper’s authors also disclosed conflicts of interest — equity in digital health companies, advisory work with Headspace and YouTube. This kind of disclosure is information readers need. But disclosure alone doesn’t mean the conclusions are correct or free of bias. The claims and the evidence need to be examined separately.

This is also the first thing I get agreement on when defining success metrics in consulting: “What would we need to see to admit we failed?” If you don’t set the criteria for failure in advance, you can call any outcome a success no matter what happens.

I believe we need to set criteria for judging success and failure before restricting teenagers’ usage. We need to look not just at time spent, but at sleep, school life, mental health, and exposure to harmful content together. And a service’s safety design and restriction policies should be something we can keep improving by comparing their actual effects.

Closing

In Australia, many accounts were restricted, and parent surveys also showed a drop in overall account-holding rates. At the same time, a significant share of existing users kept their accounts. These early results alone don’t tell us anything about mental health effects. We still need to see how teenagers’ daily lives and safety actually change going forward.

The next time you see an announcement declaring a policy or product a “success,” I’d recommend asking just one question first: “What exactly is this number measuring?” That question alone can go a long way toward separating the announced figure from the actual outcome. If you’re curious to dig deeper into the effects of the ban, I’d suggest starting with Chapter 4 of the paper below (the section proposing evaluation methods).

💬 Have you ever run into a case—at work or in daily life—where a metric touted as a “success” turned out to be at odds with reality? Tell us in the comments what the number was.

References & Further Reading

Primary sources

• Neff Lind, M., Schueller, S. M., & Odgers, C. L., “We don’t know how social media bans will affect youth but we’re doing it anyway!”, Frontiers in Developmental Psychology, 2026. : This is the backbone of today’s newsletter. Table 1, which summarizes 40 studies, and Chapter 4, which proposes an evaluation method, are the key parts.
• Burnell, K. et al., “The effects of social media restriction: meta-analytic evidence from randomized controlled trials”, SSM-Mental Health, 2025. Shows numerically why the effects of restriction are so small.
• Ferguson, C. J., “Do social media experiments prove a link with mental health”, Psychology of Popular Media, 2024. : The source for the conclusion that the average effect is indistinguishable from zero.
• Davis, C. G., & Goldfield, G. S., “Limiting social media use decreases depression, anxiety, and fear of missing out in youth with emotional distress”, Psychology of Popular Media, 2025. : Conversely, this study found benefits for a specific group of young people. Worth reading alongside the others for balance.

Background

• eSafety Commissioner, “Social Media Minimum Age: March 2026 Compliance Update”, Australian Government, 2026. : Lets you check both the retention rate among existing account holders and the overall drop in account ownership across the full survey population.
• National Academies of Sciences, Engineering, and Medicine, “Social Media and Adolescent Health”, 2024. : The U.S. National Academies’ conclusion that “the relationship is small and complicated.” Worth reading as the starting point of this whole debate.
• Jonathan Haidt, “The Anxious Generation”, 2024. : The flagship argument for the pro-ban side. Knowing the opposing view is what keeps the picture balanced. Published in Korea under the translated title Bulan Sedae (The Anxious Generation).

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Vanity metric: a figure that looks big and impressive but doesn’t actually connect to the goal you’re trying to reach. A classic example is high view counts that never translate into purchases. ↩

2. Randomized controlled trial (RCT): a method that randomly splits participants into two groups (treatment/comparison) to compare effects. Considered the strongest tool for isolating cause and effect. ↩

3. Effect size (g): a value indicating how “big” an effect is. Typically 0.2 counts as small and 0.5 as medium. 0.17 falls on the small end. ↩

4. Demand characteristics: a phenomenon where participants sense the intent of an experiment and change their behavior to match expectations. This can inflate results. ↩