16.3 Platforms as speech gatekeepers

Much of today's public speech runs through a handful of large platforms. They are not courts, yet their decisions can decide which ideas are visible, which communities can organise and which livelihoods continue.

This gatekeeping rarely looks constitutional. It arrives through terms of service, automated ranking systems and uneven paths to appeal. You do not need a legal background to understand it, but you do need to watch how the rules are written and applied in practice.

Terms of service as law

Sign up for a social network, video site, marketplace or app store and you agree to a contract: the terms of service and related policies. In the UK that is not a statute, but it works like a private rulebook with real-world consequences.

Platforms can remove posts, delete accounts, restrict features or demonetise content under those terms. Not theoretical:

• a journalist can have years of work removed from a platform archive;
• a charity can lose its fundraising tools;
• a small business can see its shop listing vanish overnight because a policy changed.

Terms mix broad values with specific bans. "Hate speech", "misinformation" and "harm" may be defined, but the definitions are deliberately flexible to cover new harms and close obvious loopholes.

That flexibility lets rules be enforced quickly. It is also why outcomes feel inconsistent: two similar posts can be treated differently because the platform's internal context is not yours.

Example: a UK nurse discussing vaccine side effects in a professional group might be flagged under a misinformation rule if the system cannot tell it apart from conspiracy content elsewhere.

Myth: a platform's rules are the same as the law. Reality: they are not.

• Something lawful in the UK may still be banned by a platform.
• Something illegal in one country may stay visible in another, because the platform applies local filtering.

The result is a patchwork where the practical boundaries of speech are set by private policy, not by Parliament or the courts.

That is not always sinister; platforms are often trying to keep services usable and safe. But everyday expectations about free expression do not transfer neatly to a privately run digital space.

Mitigations here are about predictability and resilience, not perfect protection:

• read the policies you rely on most (tedious but practical, especially if your income depends on a platform);
• keep a copy of your work offline;
• keep a presence on more than one platform;
• use your own website or mailing list.

None of this changes the power imbalance, but it reduces the damage when the rules move or are applied unexpectedly.

Algorithmic suppression

Gatekeeping is not only about what stays up. On most platforms, ranking algorithms (sets of rules and models for sorting content) decide what people see first, if at all.

No deliberate censorship is needed to suppress speech. A system tuned for engagement may quietly demote content that will not keep people scrolling, even if it is accurate, lawful and socially valuable.

Two everyday examples:

• A teacher posts a detailed thread on GCSE revision methods. Informative, but it does not spark quick reactions, so it sinks under a flood of short clips.
• A local councillor shares a sober update about roadworks. Few comments, so fewer people are shown it.

Neither was banned, but both had their reach cut. That is gatekeeping, and it happens constantly.

Suppression can also be intentional but opaque. Platforms may cut the distribution of content that brushes a policy boundary without removing it, often called "visibility filtering" or "downranking". Reasons include:

• slowing the spread of coordinated manipulation;
• limiting graphic content while keeping it for documentation;
• reducing the amplification of harassment.

The risk is that it is hard to detect from outside.

Myth: weak performance means a "shadow ban". Reality: the likelier cause is algorithmic indifference, timing or competition for attention. The idea of universal, deliberate suppression distracts from the mundane truth: most content is simply not prioritised.

Mitigations are limited, because ranking systems are proprietary and constantly tweaked. You can still:

• reduce dependence on a single feed by encouraging direct subscriptions, newsletters or RSS;
• build in redundancy for important information: mirror it elsewhere, or provide a stable, linkable source that can be shared even if the original post is downranked;
• put essential details in the first few lines or in the linked source (useful for organisations), so partial visibility still gets the message across.

Appeal asymmetry

When enforcement goes wrong, appeals matter. But they are often asymmetrical: the platform acts instantly, while you face delays, automated replies or no practical route to a human.

That is not always malice. At millions of reports a day, manual review is expensive and slow. But the asymmetry shapes behaviour: knowing a takedown may be hard to reverse, people self-censor or avoid controversial but lawful topics.

It bites hardest when accounts are tied to income or public service:

• a UK musician relying on video revenue loses months of earnings after a mistaken copyright claim;
• a community organiser loses event tools during a local emergency because their account was automatically flagged.

Even when appeals succeed, the damage is often done. You lose not just speech but time, attention and trust.

Partial mitigations:

• keep a clear archive of your original content, timestamps and licences, which makes intellectual property appeals easier;
• use verification or professional accounts where offered, for better support channels (though these often favour larger organisations);
• build relationships across platforms and keep off-platform contact methods, so a sudden restriction hurts less.

None of these guarantees a fair outcome, and that is worth saying plainly. Platforms prioritise speed and scale over due process, and you have to decide how much risk you will accept when choosing where to speak and build a community.

When the state stands behind the gatekeeper

Everything above treats gatekeeping as a largely private, commercial affair: companies enforcing their own rules for their own reasons. That picture is now incomplete.

With the Online Safety Act, examined in 17.1, the state stands behind the gatekeeper, attaching legal duties and severe financial penalties to how platforms handle content.

That shifts the incentives for everyone who speaks online. A company weighing a borderline post is no longer just managing its reputation and user experience; it is managing the risk of regulatory enforcement, and that risk reliably pushes towards removing more rather than less.

This turns private discretion into something closer to public censorship by proxy. Quick comparison:

• A platform over-removes lawful speech under its own terms: a commercial choice. You can respond by going elsewhere.
• It over-removes because the law penalises guessing wrong: the pressure is systemic, and the alternatives shrink because every comparable platform faces the same incentive.

So the patchwork of private rulebooks described earlier is increasingly underwritten by a single legal framework, and cautious, automated over-removal becomes the norm everywhere rather than one company's quirk.

This is how broad regulation, however well-intentioned, narrows the practical boundaries of speech. It is why the developments in Chapter 17 belong in any honest account of who controls what you can say.

For you, the earlier lessons hold, with added weight. The resilience measures (off-platform copies, your own website or mailing list, not depending on one service for your audience or income) matter more when state-backed pressure makes removals more likely and more uniform.

The limits on your speech are now a blend of company policy and legal compulsion, not company policy alone. Knowing that lets you set realistic expectations and choose deliberately where, and under what name, you build a public presence.