Nearly 50 million content moderation decisions have been reversed by online platforms in the EU in just two years.

‍
The figure, published by the European Commission in February 2026, offers a striking insight into how content moderation actually works on digital platforms. Since the Digital Services Act (DSA) became applicable, users have appealed more than 165 million moderation decisions through platforms’ internal complaint mechanisms. Almost 30% of those decisions were subsequently reversed.

‍
The numbers become even more interesting when looking at what platforms are actually moderating. In the first half of 2025, 99% of content moderation decisions concerned the enforcement of platforms’ own terms and conditions, rather than the removal of content reported as illegal under EU or national law.

‍
This changes the way businesses should think about content moderation.
Content moderation is not simply about identifying illegal content and taking it down. It is increasingly a governance process through which platforms decide what users can post, what remains visible, what is restricted, which accounts can continue to operate and under which conditions.

‍
From removal to platform governance
A moderation decision may involve much more than deleting a post.
Platforms may restrict or remove content, suspend or terminate an account, reduce visibility, limit access to specific functionalities, apply labels or impose other restrictions under their terms and conditions.
The DSA has therefore introduced an important procedural dimension: users must have meaningful ways to challenge moderation decisions affecting their content or accounts.

‍
And the data show why this matters.
A reversal rate of almost 30% among appealed decisions means that a substantial number of initial moderation decisions do not remain unchanged after review. The issue is therefore not only how platforms moderate, but also how they review their own decisions.

‍
The first question for platforms: what exactly are you enforcing?
The distinction between illegal-content moderation and terms-and-conditions enforcement is crucial.
A platform may remove content because it is unlawful. But it may also remove or restrict content because it violates contractual rules established by the platform itself.
These are different grounds for intervention and should be treated accordingly.
For businesses operating platforms, this raises a practical question:
Can you clearly identify the legal or contractual basis for each moderation decision?
A robust moderation framework should therefore distinguish between:
• content that is illegal;
• content that violates platform terms and conditions;
• content subject to restrictions under specific platform policies;
• decisions based on automated detection;
• decisions subsequently reviewed by a human operator.

Without this distinction, moderation data can become difficult to interpret and moderation decisions difficult to defend.

‍
A moderation decision is only the beginning
The DSA framework places increasing emphasis on what happens after a moderation decision.
Users can challenge decisions through the platform’s internal complaint mechanisms and, under certain conditions, through out-of-court dispute settlement bodies. In the first half of 2025, these bodies reviewed more than 1,800 disputes concerning content on Facebook, Instagram and TikTok and reversed the platforms’ decisions in 52% of closed cases.
This creates a feedback loop:
Detect → Decide → Explain → Appeal → Review → Learn
‍

For platforms, the last step is particularly important.
A reversal should not simply mean that one individual decision has been corrected. It can also provide information about the performance of moderation rules, automated systems, internal policies and review procedures.
What should platforms measure?
The Commission’s DSA framework increasingly requires transparency around moderation practices, including information on content removals, automated moderation systems, user complaints and moderation teams.
For businesses, this suggests a broader compliance question:
Do you have the data needed to understand whether your moderation system is actually working?
Useful indicators include:
• number and type of moderation decisions;
• legal versus contractual grounds for intervention;
• automated versus human decisions;
• complaints and appeals;
• reversal rates;
• reasons for reversals;
• time required to review decisions;
• recurring errors or problematic categories;
• performance of automated detection systems.

‍
The objective is not simply to produce a transparency report. It is to create a measurable governance system in which moderation decisions can be monitored, challenged and improved.
From content moderation to moderation governance
The EU experience shows that content moderation cannot be treated as a one-way process in which platforms identify problematic content and users simply accept the result.
The nearly 50 million reversed decisions reported by the European Commission point towards a different model: moderation must also incorporate review, accountability and learning mechanisms.

‍
For platforms, the practical question is therefore no longer simply:
“Can we remove problematic content?”
It is:
“Can we demonstrate that our moderation decisions are based on clear rules, properly explained, reviewable and capable of being corrected when they are wrong?”
That is where content moderation becomes a platform governance issue.

‍