Social media companies should not moderate political content

What's this about?

People disagree about whether social media sites should control posts about politics. The key question is how to stop clear harm without unfairly silencing people.

What supporters say

• Site rules can bend public debate by hiding some views and boosting others.
• Companies may make unfair choices when they judge posts about politics.
• Too much control can silence small groups or views that leaders dislike.
• People should often see and debate political ideas, even when those ideas seem wrong.

What critics say

• Sites can limit false election claims that spread fear and doubt about voting.
• After Jan. 6, account bans cut the reach of false claims about the U.S. election.
• Small prompts that ask users to check truth can help them share better posts.
• Labels can warn about AI-made posts or propaganda while still letting people read them.

The bottom line

The evidence does not support banning all political post rules. Sites should keep limited power to stop clear harm, with strong rules that prevent unfair silencing.

Social media companies should not moderate political content, critics argue, because such decisions can distort public debate and unfairly silence voices. But the evidence points instead to a narrower approach: platforms should retain limited powers to act against clear harms, under strong safeguards.

The case for

The strongest argument for moderation is that it can reduce the spread of false election claims. A peer-reviewed study in *Nature* found that suspending prominent accounts after the January 6 attack on the U.S. Capitol reduced the reach of election misinformation on Twitter, now X. Evidence from the 2020 election also documented how widespread and persistent such misinformation could be online 1.

That does not mean every political post should be removed. Less restrictive tools can make a difference. In a randomized study, prompting users to think about accuracy improved the quality of what they later chose to share, and the effect lasted over time 2. A broader review found that misinformation interventions can work, although their impact varies depending on the design, setting and audience.

Labels may offer another middle course. Research suggests that labels on propaganda or AI-generated material can change how people interpret or share content while leaving it available to read 3. Their success, however, depends on clear wording, whether users notice them and whether they trust the platform applying them.

The claim that any moderation of political material is inherently censorship also faces legal and institutional limits. In *Moody v. NetChoice*, the U.S. Supreme Court said that at least some platform decisions to curate or remove material may count as protected editorial judgment. Major social networks are also important places where people encounter political information, making the way they manage content more than a purely private technical matter 4.

The case against

The case for restraint is serious. Political moderation can look, or become, like partisan gatekeeping, especially when rules are vague and political language depends heavily on context. Research finds that disputes over moderation are not simply arguments about what is factually true; they also reflect deeper political conflict. It can be difficult to tell the difference between viewpoint discrimination and neutral enforcement of rules 5.

Mistakes may fall especially hard on marginalized speakers. Studies of AI moderation in the Global South have found that centralized systems can misunderstand local language and political context, offering uneven protection for freedom of expression. Other research has raised concerns about inconsistent enforcement across languages, regions and political groups 6.

Public oversight is also weak. Reviews of the European Union’s Digital Services Act transparency database have found that platform reports can be inconsistent, hard to interpret and based on companies’ own accounts. During election periods, available records have still left major questions about how moderation decisions were made and whether they were applied fairly 7.

Nor does reducing a post’s reach necessarily solve the broader problem. Deplatforming can make misinformation communities less visible while encouraging them to move elsewhere or reorganize. And reporting on a large Facebook experiment found that algorithm changes did not substantially change political polarization during the period studied (see Figure 3) 8. These findings warn against assuming that lower visibility automatically means less division or a lasting democratic benefit.

The bottom line

The evidence does not support a blanket rule that social media companies should never moderate political content. Targeted action has been shown to reduce the reach of election misinformation, while accuracy prompts and labels may curb deception without denying access altogether 12.

But platforms should not have unlimited discretion. Political speech requires heightened protection, with clear public rules, explanations, appeals and independent auditing. The strongest approach is narrow, proportionate intervention: use tougher measures only for clearly defined and serious harms, and prefer prompts or labels when claims are disputed but not plainly dangerous.

This conclusion carries high confidence, not because moderation is risk-free, but because the record shows both its benefits and its real dangers. The unresolved question is whether platforms can apply these tools consistently and transparently enough to avoid the documented risks of bias, error and unequal treatment.