TikTok removed 184 million videos globally in a single quarter, about 0.5% of everything uploaded, and caught 99.7% of violations in some markets before any user reported them. Those numbers sound like a system working well, and they may be, but they cannot be compared to anything: Meta reports prevalence as a share of views, TikTok reports removals as a share of uploads, and YouTube reports raw counts. Three platforms, three denominators, no common measure. Meanwhile Meta disclosed that a bug more than doubled its false positives during part of the quarter. The figures below come from platform transparency reports.
TL;DR
- TikTok removed 184,012,576 videos globally in Q1 2026 (TikTok)
- That was about 0.5% of all content uploaded (TikTok)
- YouTube removed 69,271,993 videos in January to March 2026 (Google)
- YouTube terminated 2,183,919 channels in the same period (Google)
- TikTok removed more than 4.8 million videos in Nigeria alone (TikTok)
- That represented 0.6% of Nigerian uploads (TikTok)
- TikTok removed 884,591 videos in Kenya (TikTok)
- 99.7% of Kenyan removals were proactive (TikTok)
- 96.3% of Kenyan removals happened within 24 hours (TikTok)
- TikTok removed 1,192,526 videos in South Africa (TikTok)
- Meta reports Facebook violent content prevalence at 0.15% to 0.16% (Meta)
- That is down from 0.19% to 0.20% (Meta)
- Meta disclosed false positives rising over 100% during a Q4 bug (Meta)
1. The Removal Volumes
Absolute counts are the easiest numbers to publish and the hardest to interpret. TikTok removed 184,012,576 videos globally in the first quarter of 2026, while YouTube removed 69,271,993 videos and terminated 2,183,919 channels over January to March. Both figures are enormous and neither tells you whether the platform is getting safer, because removals rise with enforcement effort as well as with violation volume.
| Metric | Value | Source |
|---|---|---|
| TikTok videos removed globally, Q1 2026 | 184,012,576 | TikTok |
| Share of all content uploaded | about 0.5% | TikTok |
| YouTube videos removed, Jan to Mar 2026 | 69,271,993 | |
| YouTube channels terminated | 2,183,919 | |
| Ratio of TikTok to YouTube video removals | roughly 2.7 to 1 | Derived |
| TikTok videos removed in Nigeria | more than 4.8 million | TikTok |
| Nigerian removals as a share of uploads | 0.6% | TikTok |
| TikTok videos removed in South Africa | 1,192,526 | TikTok |
| TikTok videos removed in Kenya | 884,591 | TikTok |
The consistency of the violation rate across markets is the more interesting finding: about 0.5% globally and 0.6% in Nigeria, in populations with very different content cultures.
That stability invites an uncomfortable question about what the rate is measuring. A violation rate that barely moves across markets with different languages, laws, and norms could mean human behaviour is remarkably uniform, or it could mean the detection system has a roughly fixed capacity and finds about as much as it is tuned to find. Nothing in the published data distinguishes those two explanations, because platforms do not disclose how detection thresholds are set or how they vary by market. Until they do, a removal rate is at least as much a description of the classifier as of the users, and a platform reporting a stable violation rate is not necessarily reporting a stable problem. Source: Google Transparency Report, YouTube Community Guidelines enforcement.
2. Prevalence Is the Metric That Matters
Meta reports a fundamentally different measure, and it is the better one. Prevalence estimates how much violating content people actually see as a share of views, rather than how much was removed, and Meta reported violent and graphic content on Facebook falling to 0.15% to 0.16% from 0.19% to 0.20%. Adult nudity on Instagram moved the other way, rising to 0.09% to 0.11% from 0.06% to 0.07%.
| Metric | Value | Source |
|---|---|---|
| Facebook violent and graphic content prevalence | 0.15% to 0.16% | Meta |
| Prior period | 0.19% to 0.20% | Meta |
| Direction | improved | Meta |
| Instagram adult nudity prevalence | 0.09% to 0.11% | Meta |
| Prior period | 0.06% to 0.07% | Meta |
| Direction | worsened | Meta |
| Policies tracked on Facebook | 14 | Meta |
| Policies tracked on Instagram | 12 | Meta |
| What prevalence measures | violating content as a share of views | Meta |
Prevalence can fall while removals rise, or vice versa, because they measure opposite ends of the same pipeline. A platform reporting only removals is reporting its own effort rather than its users’ experience. Source: Meta Community Standards Enforcement Report.
3. Almost Everything Is Caught Proactively
The proactive detection figures are the strongest numbers any platform publishes. TikTok reported 99.7% proactive removal in Kenya and 99.6% in South Africa, meaning fewer than one violation in 250 waited for a user to report it. Speed matches: 96.3% of Kenyan removals happened within 24 hours and 94.9% in South Africa.
| Metric | Value | Source |
|---|---|---|
| Proactive removal rate, Kenya | 99.7% | TikTok |
| Proactive removal rate, South Africa | 99.6% | TikTok |
| Removed within 24 hours, Kenya | 96.3% | TikTok |
| Removed within 24 hours, South Africa | 94.9% | TikTok |
| Videos removed, Kenya | 884,591 | TikTok |
| Videos removed, South Africa | 1,192,526 | TikTok |
| Implied user-reported share, Kenya | approx. 0.3% | Derived from TikTok figures |
| What proactive detection cannot measure | violations never detected at all | Derived |
The limitation is structural and worth stating: a proactive detection rate is calculated over content the platform found, so it says nothing about violations that were never detected by either automation or users. Harassment context sits in our online harassment statistics. Source: TikTok Q1 2026 enforcement reporting as covered regionally.
4. Errors, Including One Quantified
Enforcement mistakes are rarely quantified in public, which makes Meta’s disclosure unusual. Meta reported that its false positive count increased by over 100% during a brief period in the fourth quarter of 2025 because of a bug, while describing its overall error rate as materially unchanged. A doubling of wrongful removals, even briefly, at Meta’s scale represents a very large absolute number of affected posts.
| Metric | Value | Source |
|---|---|---|
| Increase in false positives during the incident | over 100% | Meta |
| Duration | a very brief period of Q4 2025 | Meta |
| Stated cause | a bug | Meta |
| Overall error rate characterisation | materially unchanged | Meta |
| Policies tracked on Facebook | 14 | Meta |
| Policies tracked on Instagram | 12 | Meta |
| Reporting period covered | Q4 2025, September to December | Meta |
| Frequency of comparable disclosures industry-wide | rare | Derived |
Most platforms do not publish false positive rates at all, so the absence of comparable figures elsewhere reflects disclosure practice rather than better performance. Meta deserves some credit for disclosing an embarrassing number, and the fact that doing so is unusual is itself a finding about the state of platform transparency. Source: Meta integrity reports, H1 2026.
5. The Reporting Is Getting Less Frequent
Transparency cadence moved in the wrong direction. Meta shifted its Community Standards Enforcement Report to a semi-annual schedule, so the most recent edition covered the fourth quarter of 2025 and the next covers the first half of 2026. Halving the reporting frequency on a system that changes weekly means the public record is structurally stale.
| Metric | Value | Source |
|---|---|---|
| Meta reporting cadence | semi-annual | Meta |
| Most recent period covered | Q4 2025 | Meta |
| Next period to be covered | Q1 and Q2 2026 | Meta |
| Publication timing of the next edition | second half of 2026 | Meta |
| YouTube reporting period available | January to March 2026 | |
| TikTok reporting period available | Q1 2026 | TikTok |
| Common metric across the three platforms | none | Derived |
| Denominators used | views, uploads, absolute counts | Derived |
The deeper problem is not frequency but incomparability: Meta reports against views, TikTok against uploads, YouTube against nothing at all. No regulator currently requires a common denominator.
The incentives explain the divergence. Each platform publishes the metric on which it performs best and which it controls the definition of. Removal counts make a platform look diligent. Proactive detection rates make it look technically capable. Prevalence, which measures what users actually experience, is the metric most useful to the public and the least flattering to publish, which is presumably why only one of the three reports it. Voluntary transparency will keep producing incomparable data until someone with authority specifies the denominator, and nobody currently does. Platform context sits in our social media statistics, Facebook statistics, and teen social media statistics. Source: Meta on sharing Community Standards Enforcement results.
Summary: Content Moderation by the Numbers
| Metric | Value | Source |
|---|---|---|
| TikTok videos removed globally, Q1 2026 | 184,012,576 | TikTok |
| Share of uploads removed | about 0.5% | TikTok |
| YouTube videos removed, Jan to Mar 2026 | 69,271,993 | |
| YouTube channels terminated | 2,183,919 | |
| TikTok removals, Nigeria | more than 4.8 million | TikTok |
| Nigerian removals as share of uploads | 0.6% | TikTok |
| TikTok removals, South Africa | 1,192,526 | TikTok |
| TikTok removals, Kenya | 884,591 | TikTok |
| Proactive removal rate, Kenya | 99.7% | TikTok |
| Proactive removal rate, South Africa | 99.6% | TikTok |
| Removed within 24 hours, Kenya | 96.3% | TikTok |
| Removed within 24 hours, South Africa | 94.9% | TikTok |
| Facebook violent content prevalence | 0.15% to 0.16% | Meta |
| Prior period | 0.19% to 0.20% | Meta |
| Instagram adult nudity prevalence | 0.09% to 0.11% | Meta |
| Prior period | 0.06% to 0.07% | Meta |
| Increase in false positives during the bug | over 100% | Meta |
| Policies tracked on Facebook | 14 | Meta |
| Policies tracked on Instagram | 12 | Meta |
| Meta reporting cadence | semi-annual | Meta |
Methodology and Sources
- Prevalence rates, policy counts, the false positive disclosure, and reporting cadence come from Meta’s Community Standards Enforcement Report and integrity reporting (Community Standards Enforcement Report, integrity reports H1 2026, how Meta shares enforcement results, Transparency Center).
- YouTube removal and channel termination counts come from the Google Transparency Report (YouTube Community Guidelines enforcement, methodology notes).
- TikTok global and market-level removal volumes, proactive detection rates, and removal speed come from its Q1 2026 Community Guidelines Enforcement Report as covered in regional reporting (Techweez on Kenya, Vanguard on Nigeria, SME Tech Guru on South Africa).
- Data watch: figures from these three platforms cannot be compared. Meta reports prevalence as a share of content views, TikTok reports removals as a share of uploads, and YouTube reports absolute counts with no denominator, so no like-for-like ranking is possible and none should be inferred. All figures are self-reported by the platforms with no independent audit; no regulator verifies them and no common standard governs what must be disclosed. Proactive detection rates are calculated over content the platform found, so they cannot account for violations never detected by either automation or users, which means a high proactive rate is consistent with poor overall coverage. Meta’s semi-annual cadence means its most recent published data covers late 2025. Rows marked as derived are arithmetic or direct inference from published figures.
- Last updated: August 2, 2026. We update this roundup quarterly as platforms publish new transparency reports.