Meta’s $18 Billion Settlement Is a Beginning, Not the Answer: The Real Threat to Democracy Is the Algorithm
The $18 billion settlement agreed by Meta on August 25th in a landmark case brought in California by 48 state attorneys-general and other plaintiffs is being described in many quarters as a turning point. The sum is staggering, yet, as Clare Melford notes, the money is trivial when set against Meta’s enormous profits. The true significance of the settlement lies elsewhere: in the company’s long-overdue pledge to introduce controls over how and when children use its platforms. This commitment, however modest, has also placed competitive pressure on the other two dominant firms in the social media business, Alphabet and TikTok, to follow suit. But to regard this as a genuine solution to the crisis of the information ecosystem would be a profound mistake. The settlement addresses only the visible surface of a much deeper structural problem. While restricting children’s access to social media is a necessary and humane step, it leaves untouched the central engine of the modern public sphere: the recommendation algorithms that determine what billions of people see, hear, and believe. These systems, designed by a handful of corporations to maximise engagement and advertising revenue, are the true architects of our shared reality, and they operate with almost no legal accountability, no public-interest mandate, and no meaningful oversight. The Meta settlement may be a welcome start, but it is only a start, and the hardest and most consequential work lies ahead.
To understand why this is the case, one must step back and consider the wider picture. Trust in news is falling across nearly every democracy. Disinformation spreads with extraordinary speed, often outpacing the truth by orders of magnitude. Artificial intelligence is already changing how people search for information and, increasingly, how they consume it. Traditional journalism is under severe economic pressure, with newsroom closures, layoffs, and the consolidation of media ownership all pointing to an uncertain future. Meanwhile, concerns about political polarisation, online toxicity, and the erosion of shared factual ground show little sign of fading. These developments are usually discussed as separate problems, each requiring its own technical or regulatory solution. Fact-checking services, content labelling, media literacy campaigns, and tougher moderation policies are all proposed as remedies. All of these efforts are important in their own right, but they share a common and deeply flawed assumption: that improving the information ecosystem starts with improving the information itself. This is to look at the problem from the wrong end. In the internet age, focusing on content in isolation is like telling every household to boil its drinking water in order to deal with the unhealthy stuff coming out of the pipes. It places the burden on the consumer while leaving the source of contamination untouched. The water will remain polluted no matter how diligently individual households boil it, and the same is true of our information environment.
The defining feature of today’s information ecosystem is not the content that we, as users, produce. It is the system that determines what content actually reaches us. Algorithms are not neutral conduits; they are editorial machines, making millions of decisions every second about what is relevant, what is urgent, and what deserves attention. They are designed to optimise for engagement, and engagement is not the same as truth, accuracy, or public value. The most shocking, frightening, and emotionally charged material tends to perform best because it triggers the strongest responses, and the algorithms have learned to feed on these responses with ruthless efficiency. If we want to understand why the information ecosystem looks the way it does, or how it might be improved, we need to spend far less time thinking about information in isolation and far more time thinking about the algorithms that mediate our relationship with it. And we need to give the platforms that design and operate these algorithms the same sort of legal responsibilities that we have for decades given to publishers. The platforms do not produce the content themselves, but they decide what content you see, how it is ranked, and how it is amplified. They shape the agenda, set the tone, and control the flow of public discourse. In any meaningful sense, they are publishers, and they should bear the responsibilities that come with that role.
The scale of this transformation is difficult to overstate. The internet has become the information ecosystem. The average person now spends almost seven hours a day on an internet-connected device, compared with roughly five hours spent on radio, print, and television combined. As the eyeballs move online, so too do the advertising dollars that pay for the consumer internet. Online advertising now accounts for more than $700 billion a year, while all other media combined attract only about $200 billion. That economic shift has profound consequences. Those seven hours are being spent disproportionately on the products of just three companies: Alphabet, Meta, and TikTok. About 3.8 hours per day are spent on their respective platforms, YouTube, Instagram, and TikTok. What those eyeballs are doing in those hours has also changed. We used to talk about “browsing the web,” a phrase that implied agency, curiosity, and intentionality. We were active participants, searching for things in a vast library. But now, by some measures, 70 percent of those hours spent on social media count as passive consumption rather than the making of deliberate choices. People are not choosing what to see; they are simply watching what the platforms feed to them. These curated algorithms are known as UpNext on YouTube and For You on TikTok. Meta does not publish figures for its Facebook and Instagram algorithms, but the pattern is the same. The feed is not a neutral stream of content; it is a personalised, automated, and continuously optimised sequence of recommendations, designed to keep users hooked and to sell their attention to advertisers.
Nowhere is the consequence of this shift more visible than in the changing habits of news consumption. In the past, news publishers, think tanks, and policy circles could argue that this flow of information was somehow separate from “the news,” which was still primarily consumed in a traditional way on television and in print. No longer. In the United States, 54 percent of people now access news through social media. The 2026 edition of the Digital News Report from the Reuters Institute for the Study of Journalism at Oxford revealed that, in the 48 countries surveyed, social media and video networks have, for the first time, overtaken television and owned websites as the most popular source of news. This represents one of the most profound changes in the history of media. Editors have not disappeared; they have been replaced by machine learning systems that optimise billions of personalised recommendations every day to keep us engaged and to show us adverts. Most of these systems are under the ultimate control of just three companies, none of which has a public service mandate. They are not accountable to any electorate, they are not bound by journalistic ethics, and they are not subject to the professional standards that govern traditional newsrooms. Their only obligation, in practice, is to their shareholders, and the most reliable way to serve those shareholders is to keep people scrolling, clicking, and reacting, regardless of the social cost.
If we want to improve the information ecosystem, we must stop complaining about the content within it and start applying standards to the algorithms that curate it for us. Setting controls over how and when children access social media platforms is useful, but it does not address the real issue, for children or for adults. The real issue is how the flow of information and news is determined, and in particular how the algorithms favour content that is shocking, frightening, or emotional because doing so is profitable. Crucially, applying standards to algorithms is not akin to censorship. It will not seek to determine what content is created, but rather the means by which it is distributed and amplified. This can be done by governments setting requirements for algorithmic transparency, independent audits, and content assessments. It can be done by requiring users to be given a genuine means by which they can choose how their feeds are ranked, rather than being subjected to a hidden, corporate-defined formula. It can be done by enabling users to move their accounts and social graphs between platforms, thereby breaking the stranglehold of the three dominant companies and creating room for competition. These are not radical or authoritarian measures. They are the basic regulatory protections that we apply to industries that have a profound impact on public life. The internet is no longer a niche technology; it is the public square, the library, the newsstand, and the town hall all rolled into one. The question is whether we are prepared to treat it as such. For too long, we have told people to boil their tap water, to be careful about what they read, to fact-check, to be sceptical, to protect themselves from the toxic flows of the algorithmic pipe. It is time to tell the water companies to put safer water into the pipes in the first place. The Meta settlement is a small and welcome step in that direction, but the real battle lies in the engine-room of the internet, and it is a battle we have only just begun.