Can banning the endless scroll make children safer?

The next item is already waiting. Another video begins, another set of recommendations appears, and a stopping point becomes something the user has to create rather than something the product supplies.

California’s AB 1709 takes that design question into regulation. Signed on 10 September, the law targets specified addictive features on covered platforms for users under 16. Its sponsor, Common Sense Media, describes restrictions on personalised feeds and autoplay, with the law taking effect on 1 January 2027. The signing is not evidence that those changes have already produced a health benefit. [1][2]

The policy shift is significant. Instead of asking only which posts should be removed, it asks how the service structures the experience of using it. That opens a practical debate about safety, age assurance, access and what an effective result would look like.

A feature is not a diagnosis

“Addictive” has a specific role in the legislation’s definitions. It should not be treated as a clinical diagnosis of every young person who uses a feed. The covered features and the scope of the platform rules need to be read in the actual law and subsequent guidance. [2]

The case for addressing design is that choices about repetition, recommendation and automatic continuation can influence how an experience unfolds. A product can make stopping more or less deliberate. But the effect of a particular design on a particular person is not established merely by describing the feature.

That distinction is reflected in the research debate. The National Academies’ assessment of social media and adolescent health found that the evidence did not support a simple population-wide causal verdict. It identified the potential for both harm and benefit, difficulties in establishing the direction of relationships, and a need for better understanding of particular features. [3]

The appropriate conclusion is not that nothing can be done until every uncertainty disappears. It is that a policy should state the harm it aims to reduce, the mechanism it changes and the evidence that would show whether it worked.

Knowing the user’s age is part of the problem

An age-based rule requires a way to identify which experience a user should receive. That creates questions beyond the design of the feed itself. What information is required? How accurate is the process? How can a mistake be corrected? What happens to information collected for the check?

Supporters describe AB 1709 as requiring reasonable age-related measures and limiting specified features rather than necessarily removing all younger users from a service. Its sponsor also identifies an advisory commission and an enforcement role for public authorities. [2]

The implementation will need to make those distinctions real. A protective system that is difficult to use can exclude someone incorrectly. A check that gathers more information than necessary can create a separate privacy concern. A process that relies on a label without verifying its meaning may fail to deliver the intended distinction between users.

These are design and governance questions, not reasons to assume the law cannot work. They are also reasons to avoid describing age assurance as a single technical switch that resolves the policy.

Count outcomes, not just settings

Turning off a feature is an observable product change. Whether it reduces the relevant harm is a further question.

An assessment should therefore identify what changed for the people affected, not only whether a platform renamed an interface or added a warning. It should also examine unintended effects and differences among users. A shorter session, for example, would not by itself prove an improvement in every aspect of well-being; nor would the absence of an immediate population-wide change prove that no user benefited.

The National Academies’ research discussion underlines why this is difficult: online experience and health can influence each other, and individual circumstances differ. A useful evaluation needs more than a before-and-after headline. [4]

The strongest case for regulation is that public authorities can require attention to risks that a product’s ordinary incentives may not adequately address. The strongest implementation challenge is ensuring that the chosen restrictions map onto evidence rather than simply sounding protective.

Both should remain in the story. Parents’ concerns do not become less serious because causation is complicated. Researchers’ qualifications do not become less important because the political demand for action is strong.

The scroll is only one part of the experience

The new law sits within a wider California package that also addresses other forms of online and AI-related risk. Those measures are separate instruments with their own scope and timing. A headline about one should not be used to imply that all of them apply in the same way from the same date. [1]

For AB 1709, the next test is the transition from a signed measure to an operational change that can be evaluated. The public should be able to distinguish compliance activity, enforcement action and evidence of benefit.

Removing an automatic next step may create a stopping point. Making that stopping point part of a safer online experience will require more than declaring that the endless scroll has reached its end.