Meta has agreed that teenagers should spend no more than two hours a day on Facebook and Instagram. This is encouraging. Not merely because younger users may benefit, but because it establishes the important principle that somebody using Meta’s platforms might occasionally be told they have had enough. The next group requiring intervention is the media industry.
On 26 August 2026, Meta agreed to pay up to $18 billion and make substantial changes to Facebook and Instagram, settling claims brought by most US states and territories that it designed its platforms to encourage addictive use among children, misled consumers about their safety and improperly collected children’s data. The settlement does not establish that Meta acted wrongfully (the company continues to deny that) but demonstrates that the design choices behind engagement are now being challenged through litigation, legislation and public policy rather than accepted as immutable features of the internet.
The agreement still requires court approval. Even so, it brings to an end a closely watched federal trial at which Mark Zuckerberg had been expected to testify. We are thankfully spared the spectacle of a technology billionaire explaining human behaviour like somebody who has recently had it described to them.
The settlement requires Meta, for the next decade, to limit teenagers’ use of Facebook and Instagram to two hours a day and prevent access between midnight and 6am without parental consent. It also introduces restrictions on notifications during school hours, stronger age-assurance measures, tighter controls on age-inappropriate material, independent oversight and what the agreement calls “productive pauses”.
A productive pause sounds less like a product-safety mechanism than something suggested by Human Resources shortly before removing the biscuits from a meeting room. Still, it represents a formal attempt to place limits on the unlimited engagement upon which the social-media business model was built. Meta has, after many profitable years, discovered the off switch.
The financial figure sounds appropriately historic until it is placed alongside Meta’s approximately $201 billion in revenue during 2025. Reuters calculates that the settlement represents roughly three to four months of Meta’s profit. This makes it considerably more substantial than an ordinary corporate parking ticket, although unlikely to require a whip-round at Menlo Park or the sale of any emergency hoodies. The more consequential part of the agreement may therefore be the restrictions placed on the products themselves.
Everybody Had a Different Name for the Same Thing
The behaviour at the centre of the legal action (designing products to maximise time spent, emotional reaction and repeat visits) was never described the same way twice, because the modern digital economy relies heavily on finding reassuring names for things people might otherwise object to.
Platforms called it engagement. Newsrooms called it audience development. Advertisers called it attention. Consultants placed it inside a circular diagram and called it an ecosystem. Users, on reflection, were inclined to call it three hours of their life they would like back.
The terminology shifted with the audience, but the mechanism remained consistent: keep people on the platform, keep them reacting, and measure success by how much of their time could be absorbed before they noticed. If they became angry, anxious or unable to leave, this was not traditionally regarded as a design problem. It was an excellent session duration.
Publishers were not innocent bystanders dragged reluctantly into this system. They arrived carrying headlines and asking where to plug them in.
Throughout the 2010s, editorial decisions that had once been guided by news judgement, reader loyalty and the occasional terrifying editor-in-chief were increasingly guided by whatever the algorithm appeared to enjoy that morning. Headlines were restructured for shareability. Formats were rebuilt for the Facebook feed. Video was declared the future with such regularity that large numbers of journalists were dismissed to make room for it.
Commissioning meetings began to feature phrases such as “this could do well on social”, as though this were a journalistic criterion rather than a mild form of editorial surrender. Entire publishing strategies were based on attempting to interpret the moods of a mathematical system owned by another company. Medieval villagers studying the flight of birds had a broadly similar methodology, although rather less venture capital.
The media industry did not merely use the platforms. It trained itself to serve them, reorganising editorial behaviour around a reward system it could neither see nor control. It was less a partnership than a very long audition for a part Meta had not promised to cast.
The Productive Pause the Media Industry Never Received
Consider what Meta’s proposed protections might look like if applied to publishers rather than teenagers.
After fifteen minutes spent refreshing the analytics dashboard, an editor would receive a gentle notification asking whether they had considered commissioning something people might actually remember. After thirty minutes, the dashboard would be replaced by a photograph of a tree and the words: “There is a world beyond concurrent users.”
A responsible adult, perhaps a sub-editor of sound judgement and adequate caffeine levels, would have to approve any headline containing SLAMS, DESTROYS, BREAKS SILENCE, FURY or YOU WON’T BELIEVE. Headlines beginning “The internet is obsessed with…” would require documentary evidence that at least seven people had noticed.
Access to real-time audience data would be blocked after midnight without a parental PIN, on the reasonable grounds that nothing good has ever been decided at 1am in response to a traffic spike. Push notifications would be disabled during working hours so journalists could concentrate on journalism. Before publishing a fourth version of the same story in a single news cycle, the system would pause to ask: “Are you building an audience, or displaying signs of platform dependency?”
Editors would be limited to two hours a day of saying “the algorithm”. Once the allowance had been exhausted, they would have to identify an actual reader and explain what that person might want.
None of this would entirely cure the problem. Some publishers would immediately open a second account and claim it belonged to their parents. But it might at least introduce a moment of reflection into an industry that has spent years responding to declining social reach by posting more frequently.
The central difficulty was always that publishers reorganised themselves around engagement signals Meta controlled and could change whenever it wished. From around 2018 onwards, Meta progressively reduced the prominence of news in its feeds. Referral traffic fell sharply. Publishers discovered that the distribution infrastructure around which they had redesigned their businesses was not, technically speaking, theirs.
The industry had trained for an algorithm that subsequently lost interest. It was like spending years learning the tango only to discover that your partner had become very interested in short-form video and was now dancing with somebody called Creator Economy.
Meta moved on. Many newsrooms are still standing beside the dance floor holding a carefully optimised headline.
Engagement Acquires a Health Warning
The more significant consequence of the settlement may not be the payment, however gratifyingly large the number appears in a press release, but the product changes it requires. If daily limits, notification controls, age assurance and independent oversight become standard safeguards, unlimited engagement begins to look less like an inevitable feature of the internet and more like a deliberate commercial choice.
That reframing matters because engagement has long been treated as a neutral indicator of public enthusiasm. More engagement meant a successful story, a healthier platform, or an especially impressive social media manager. Little distinction was made between a reader thoughtfully absorbing an investigation and somebody furiously typing beneath a photograph of a celebrity they had never met.
The dashboard counted both. Indeed, it often preferred the furious person.
Engagement is not necessarily evidence that something has informed, delighted or enriched its audience. It proves that something happened on a screen. A car alarm also generates engagement, but nobody builds an editorial strategy around one.
The settlement does not establish that Meta acted wrongfully, and the company continues to dispute the allegations. It does, however, confirm that the mechanisms used to maximise engagement deserve scrutiny rather than reverence. The algorithm is not weather. It was designed by people, adjusted by people and defended by people who generally preferred everyone else to discuss it as though it had formed naturally over the Pacific.
Publishers May Now Leave the Platform
For publishers, the lesson is not that social media should be abandoned. That is the kind of position that sounds magnificent at a conference and survives until somebody checks the traffic figures.
Social platforms remain useful for discovery, conversation and distribution. But rented distribution should not be mistaken for an owned audience, particularly when the landlord is free to remove the staircase without warning.
Newsletters, subscriptions, memberships and direct reader relationships are not glamorous substitutes for a viral post. They are infrastructure: connections that do not disappear because a platform adjusts its priorities on a Tuesday afternoon, discovers a new format or decides that news has become inconvenient.
The media spent years treating social reach as an audience when it was really temporary permission to stand in somebody else’s doorway. Meta’s settlement should remind publishers that a large number on a dashboard is not the same thing as a durable relationship, and visibility granted by an algorithm is not independence.
Teenagers may soon receive productive pauses, overnight controls and daily limits. Publishers will remain free to refresh the dashboard until their eyes water, publish a fifth version of the same story and announce that the audience strategy is “performing strongly” because somebody in Wolverhampton has posted an angry emoji.
There are, apparently, limits to how much protection society can provide.
Attention is something a platform extracts. A relationship is something a publisher earns.
This article was created, edited, illustrated, published, distributed and promoted using LettsNews, with every editorial decision remaining in the hands of its author.