Meta and TikTok both publish quarterly transparency reports on content moderation, but the two companies measure different things: Meta's Community Standards Enforcement Report centers on "prevalence" — the estimated share of views that were violating content — while TikTok's Digital Services Act report centers on volume, reporting roughly 112 million pieces of content removed across videos, live streams, ads, product listings and comments between July and December 2025. A reader who tries to line the two figures up side by side is comparing metrics that were never designed to match.
What do Meta and TikTok actually report?
Meta's Transparency Center defines five core metrics in its Community Standards Enforcement Report: prevalence, "the number of times people see violating content"; content actioned, the count of posts, photos, videos or comments removed or labeled; proactive rate, the share of actioned content Meta says it found before users reported it; appeals, the volume of enforcement decisions users contested; and restorations, the content reinstated after appeal or review.
TikTok's sixth Digital Services Act transparency report, covering July–December 2025 and published for its roughly 178 million monthly active users in the European Union, is structured around a different set of numbers: total content removed, an automation rate of 93.8 percent for violative content actioned without human review, and an automated-decision accuracy figure of 97.6 percent. That edition also added, for the first time, enforcement volumes broken out for comments — a category the company had not previously reported separately.
Neither company's headline figure is wrong. Prevalence answers "how much of what people actually saw was violating." A removal count answers "how much did the company take down." Those are different questions, and a platform can improve on one while the other stays flat or worsens.
| Company | Headline metric | What it measures |
|---|---|---|
| Meta | Prevalence | Estimated share of content views that were violating, plus a proactive-detection rate |
| TikTok | Automation rate | Share of violative content removed without human review (93.8% in the sixth DSA report), paired with an accuracy figure |
| YouTube | Violative view rate | Estimated share of views on policy-violating video content, per the New America tracking tool's review of the platform's own disclosures |
Why doesn't a prevalence figure compare to a removal count?
The New America Open Technology Institute's Transparency Report Tracking Tool, a multi-platform review by researchers Spandana Singh and Leila Doty, found "a lack of standardization when it comes to the metrics" platforms disclose about enforcement. Facebook and Instagram's prevalence metric has no direct counterpart at most other companies; YouTube instead reports a "violative view rate," a similarly named but separately calculated figure; Reddit's reporting is organized around who removed content — volunteer moderators, platform administrators or automated systems — rather than around a single volume or rate; and TikTok's disclosures have historically emphasized underage-user and fake-account removals alongside general content-removal counts.
The researchers also flagged a category problem that outlasts any single report: some platforms group distinct violation types into single buckets. YouTube's own reporting, the tracking tool noted, lumps "spam and misleading content" together, which obscures how much of that total is misinformation specifically versus unrelated spam. A reader trying to track platform performance on a single issue — election misinformation, for instance — often cannot isolate it inside the published categories.
What do these reports leave out?
The tracking tool's authors made a pointed observation about incentives: platforms "exercise discretion over which metrics appear," and that discretion can mean publishing the numbers that read favorably while omitting ones that would complicate the story. A report is, after all, self-authored and self-audited; none of the figures cited above come from an outside verification process.
Mark MacCarthy, a nonresident senior fellow at the Brookings Institution's Center for Technology Innovation, argued in a February 2021 analysis that this gap is the core problem with voluntary transparency reporting. He proposed that platforms should disclose content moderation rules and appeal procedures, publish aggregate operational statistics, give regulators and vetted researchers algorithmic access to assess how content is prioritized, and maintain public advertising repositories. MacCarthy's broader case was that platform transparency should function the way financial-institution disclosure does — verified against a common standard, not simply self-reported — and he called for an independent audit regime rather than reliance on each company's own report.
That is largely still how it works. Meta explains what prevalence and proactive rate mean; TikTok explains its automation and accuracy rates. Neither report is independently audited against the other's framework, and no regulator currently requires a shared metric set that would let a reader compare enforcement outcomes company to company rather than company to itself over time.
What would make the numbers comparable?
Three changes recur across the record examined here. The first is a shared baseline metric — something closer to Meta's prevalence, which measures what users actually encountered rather than what a company removed, since removal totals rise and fall with how aggressively a platform enforces rules, independent of how much violating content exists. The second is consistent category definitions, so that "misinformation" or "hate speech" means the same slice of content across companies rather than being folded into broader buckets as YouTube's spam-and-misleading category was found to do. The third, per MacCarthy's Brookings analysis, is independent verification: an outside audit of the numbers a platform publishes, rather than trust in the platform's own methodology notes.
None of those three changes has been adopted uniformly. Until they are, a reader comparing Meta's prevalence rate to TikTok's automation rate, or either to YouTube's violative view rate, is comparing three different instruments rather than three readings of the same thing.
For a related policy perspective, read What does an FTC consent decree actually require a platform to do?.