Moderation decisions by the major platforms —content removals, account suspensions, restrictions— that were appealed and reviewed by an oversight body, with the focus on whether the original decision was upheld or reversed. The main source is Meta's Oversight Board, whose rulings are binding for the specific case, complemented by other appeal mechanisms. Each record documents what content was moderated, under which policy, what the reviewing body ruled (uphold, overturn, mixed) and what policy-change recommendation followed. The value of the tracker lies in measuring how often the platform's first decision turns out to be wrong when reviewed independently, and how binding that review is in practice.
Statistical readings derived from the attributes of each recorded case. All figures come from the documented events; amounts are computed only over cases with a sum expressed in the indicated currency, without converting between currencies.
Choropleth by number of forensically or judicially documented cases. Countries with no verifiable public cases remain in the base colour — the absence of events does not equal the absence of surveillance. Hover or click a coloured country to see the cases.
When a moderation decision reaches independent review, it was most likely wrong: both Meta's Oversight Board and the new EU DSA appeals body overturn around 75-80% of the cases they examine. This tracker gathers the most significant decisions and their outcome, from the Trump ban to the breast-cancer post erased by an algorithm.
Every day, the major platforms make millions of moderation decisions —what stays, what is deleted, who is suspended— and almost all are invisible and unappealable. In 2020, Meta created a body to review some of them: the Oversight Board, dubbed by the press its 'Supreme Court'. This tracker follows its most significant decisions and those of other appeal mechanisms, with a simple question behind it: when someone independent reviews the platform's decision, how often was it wrong?
The answer is uncomfortable. According to the Board's own reports, in its first year it received more than 1.1 million appeals and overturned Meta's decisions in around 70% of the cases it examined; over time, across the cumulative published decisions, that reversal rate hovers around 80%. It is not a random sample —the Board picks hard, emblematic cases— but the pattern is telling: the platform's first decision, made at scale and often by automated systems, fails far more than its silent volume suggests.
Most of the decisions reviewed in this tracker ended up overturned. The most recurrent error is not censoring a dictator, but deleting the legitimate: a breast-cancer awareness post removed by an algorithm, photos of trans people, a debate on the religious use of ayahuasca, a news report on Afghanistan or Gaza.
The tracker's decisions reveal that platforms err in both directions. The most frequent is over-zeal: removing legitimate content, almost always through automated, context-free enforcement of rules like nudity or 'dangerous individuals and organizations', the latter criticised again and again by the Board for its vagueness. The breast-cancer case, decided among the first five, is the archetype: the rule itself had a medical exception the algorithm ignored.
The opposite direction —leaving up content that should have been removed— appears less but exists, and in serious cases: in 2025 the Board overturned several Meta decisions to keep up posts shared during the 2024 UK riots that fuelled hatred. And then there is the in-between category, the Trump ban: the Board backed the suspension after the Capitol attack but rebuked Meta for imposing it in an 'indefinite and standardless' way, an invented penalty not in its rules. Upholding the decision and censuring the procedure at once neatly sums up the body's role.
Here is the nuance the tracker is careful to record: the Board's decisions are binding only for the specific case. If it orders a post restored, Meta must restore it. But its policy recommendations —changing a rule, more transparency, improving the notice system— are not mandatory. Meta accepts them, accepts them in part or rejects them, case by case. The body can correct an individual error, but it cannot force the company to change the machinery that produced it.
That is the underlying tension. An appeals mechanism that overturns 80% of what it reviews demonstrates two things at once: that a real avenue exists to correct specific abuses, and that the system generating those abuses remains basically intact, because only a tiny fraction of the millions of daily decisions ever reaches review. The Board is a window, not a brake.
For years, the only appeals instance with weight was Meta's Board, with an obvious objection: it was created and funded by the very company it judges. That changed in November 2024 with the Appeals Centre Europe, the first out-of-court dispute settlement body certified under Article 21 of the EU Digital Services Act (DSA). Unlike the Board, it covers several platforms —Facebook, Instagram, TikTok, YouTube, Threads and Pinterest— and arises from a legal obligation, not a company's goodwill.
That is the underlying tension. Appeal mechanisms that overturn around three in four of the cases they review demonstrate two things at once: that a real avenue exists to correct specific abuses, and that the system generating those abuses remains basically intact, because only a tiny fraction of the millions of daily decisions ever reaches review. They are a window, not a brake.
The Board issued in 2024 an expedited decision on two cases concerning post-election violence in Venezuela, in the context of the disputed presidential elections. Expedited decisions are an exceptional mechanism for crisis situations where the moderated content is time-sensitive. The case is part of the body's focus on 2024 as the biggest election year in history.
In April 2025, the Board ruled on three cases about posts shared during the UK riots of summer 2024 and overturned Meta's decisions to keep that content up, finding that some fuelled hatred and the risk of harm. They were the first decisions reflecting the policy and enforcement changes Meta announced in January 2025.
Meta's Oversight Board upheld in May 2021 the decision to suspend the Facebook and Instagram accounts of then former president Donald Trump after the 6 January Capitol attack, finding the restriction justified. However, it criticised Meta for imposing an 'indefinite and standardless' suspension, a penalty not in its rules, and ordered it to review the measure to set a proportionate, time-bound response. It is the body's most famous decision.
The Board overturned in 2022 Meta's decision to remove a post of a news article reporting on a Taliban spokesperson's statements about reopening schools for women and girls. The removal, under the Dangerous Individuals and Organizations policy, restricted the dissemination of newsworthy information, the body found.
In a policy advisory opinion, the Board concluded that the way Meta treated the Arabic word 'shaheed' ('martyr') caused the mass removal of content that did not incite violence, disproportionately affecting users in Arabic-speaking regions. It recommended changing the approach, in one of its broadest criticisms of the Dangerous Individuals and Organizations policy. Meta partially revised its treatment of the term.
The Board selected a case appealed by a Facebook user about a video showing an AI-manipulated version of former Brazilian footballer Ronaldo Nazário encouraging people to download an online game. The case is part of the body's focus on moderation of AI-generated content, the subject of one of its 2024 white papers on Meta's automated systems.
The Board overturned in July 2021 Meta's decision to remove an Instagram post encouraging discussion of the solitary confinement of Abdullah Öcalan, founder of the PKK. The body concluded that the removal, based on the Dangerous Individuals and Organizations policy, unduly restricted legitimate political debate, and reiterated its recurring criticism of that rule's vagueness.
The Board overturned in June 2023 Meta's decision to keep on Facebook a video in which Cambodia's then prime minister, Hun Sen, threatened his political opponents with violence ('either the legal route or a stick'). The body found Meta wrong to shield it under its newsworthiness allowance, as it clearly violated the Violence and Incitement policy, and —for the first time regarding a sitting head of government— called for suspending his Facebook and Instagram accounts for at least six months. Hun Sen reacted by urging his followers to move to TikTok and Telegram.
The Board overturned in January 2023 Meta's decision to remove two Instagram posts showing bare-chested transgender and non-binary people, taken down by automated systems under the nudity and sexual-solicitation policies. Meta acknowledged it was an enforcement error and restored the content. The Board called for clear, non-discriminatory criteria on the basis of sex or gender identity across the entire nudity policy.
In a 2024 case, the Board reversed Meta's decisions to take down a Channel 4 News (UK) report on the killing of a Palestinian child. Prompted by a Board recommendation, Meta introduced a crisis protocol to protect users' rights in high-pressure situations and preserve information of public interest.
The Oversight Board examined three posts containing the phrase 'From the River to the Sea' and upheld Meta's decision not to remove them, concluding that the phrase, on its own and in those contexts, did not violate the rules on hateful conduct, violence or dangerous individuals and organizations. The decision, sensitive given the Middle East conflict context, illustrates cases where the body backs the platform against removal requests.
The Board overturned in September 2021 Meta's decision to remove a video of a protest in Colombia in which protesters could be heard criticising then president Iván Duque and calling him 'marica'. Meta had classified the term as a slur banned under its hateful-conduct policy. The Board concluded that, in the context of political protest, the removal unduly restricted expression and recommended a public-interest exception.
In one of its first five decisions, the Board overturned in January 2021 the removal of a breast-cancer awareness post that an automated system had deleted for showing female nipples, applying the nudity rule. The Board noted that the rule itself contained an explicit exception for medical and educational purposes, and that the error revealed the limits of automated moderation without context.
The Board overturned in June 2023 Meta's decision to keep on Facebook a video inciting people to storm government buildings to protest the result of Brazil's 2022 presidential election, in the context that led to the 8 January 2023 storming in Brasília. The body concluded the content violated the Violence and Incitement policy and that keeping it up was incompatible with the company's human rights responsibilities.
The Board overturned in 2021 the removal of a post by a Brazilian spiritual school describing ayahuasca, a brew of religious and ceremonial use among Indigenous peoples. It determined the content did not violate the rules as written and recommended Meta allow discussion of traditional or religious uses of non-medical substances.
The same report identified YouTube as the least cooperative platform: of 343 in-scope YouTube matters, the Appeals Centre Europe could only resolve 29 due to missing data, as the platform did not provide the content in eligible cases. When a platform withholds information, the body usually rules for the user, finding that effective access to the DSA appeal mechanism is denied. YouTube argued the centre lacks privacy safeguards and pointed to its own internal appeals system.
The Appeals Centre Europe's first transparency report, published in September 2025 and covering November 2024 to August 2025, revealed it received nearly 10,000 disputes, of which more than 3,300 fell within its scope, and issued more than 1,500 decisions. Of these, more than three-quarters overturned decisions by platforms such as Facebook, Instagram, TikTok and YouTube, most recommending restoring content or accounts. The most frequent disputes concerned nudity, harassment and restricted goods, and covered content in 47 languages across the 27 EU member states.
In its first decisions, published in March 2025, the Appeals Centre Europe ruled for the user in 77 of 141 disputes concerning Facebook content, overturning Meta's initial decision to keep or remove the content. 76% of disputes received in that phase concerned Facebook, 21% TikTok and 3% YouTube. Meta began implementing changes based on these decisions despite their non-binding nature.
In November 2024 the Appeals Centre Europe (ACE) began operating, an independent out-of-court dispute settlement body certified by Ireland's Coimisiún na Meán under Article 21 of the Digital Services Act (DSA). It lets any EU user appeal moderation decisions —removals, suspensions or the refusal to remove— by Facebook, Instagram, TikTok, Threads, Pinterest and YouTube before independent human reviewers. It was created with a one-time grant from Meta's Oversight Board Trust, and is led by Thomas Hughes, former director of the Board itself. Its decisions are non-binding, but platforms are required to engage.
The tracker records platform moderation decisions that were submitted to an independent or quasi-independent appeals body with a public ruling. For each case it documents the platform, the type of content, the policy applied, the direction of the original decision (removal, suspension, retention) and the outcome of the review (upheld, overturned, mixed). It does not assess whether the decision is substantively correct; it documents the fact of the review and its result with a verifiable source. Meta Oversight Board rulings are binding for the specific case but its policy recommendations are not, a distinction the tracker records.