Content actioned and prevalence are not the same Meta metric

The test

Meta's Transparency Center publishes a Community Standards Enforcement Report on a semi-annual cycle, with data broken out by quarter, covering Facebook and Instagram together. The edition current as of 16 September 2026 is the H1 2026 report, which states it shares 'updated metrics for Q1 - Q2 2026' and reports that global enforcement precision, a measure of over-enforcement, averaged 'around 91% on Facebook and around 92% on Instagram.' The report tracks content broadly across the two platforms; it does not publish a metric specific to video or to Reels alone.

What the evidence says

Meta discloses two separate metrics that are easy to conflate. 'Content actioned' is a volume count of posts, photos, videos, comments or accounts that Meta removed, covered or disabled, and Meta's own methodology page warns that this number 'can go up or down due to external factors that are out of our control,' citing a hypothetical spam attack that would spike the count without reflecting better detection. 'Prevalence,' by contrast, is a sampled estimate of the share of content views across Facebook and Instagram that showed violating material, expressed as a range with a stated 95% confidence window. The H1 2026 report gives both kinds of figures side by side, for example reporting that violent-and-graphic-content prevalence rose on Facebook from a range of 0.15%–0.16% in Q4 2025 to 0.20%–0.21% in Q2 2026.

The sample and the variance

Meta's prevalence documentation states that the metric covers 'surfaces that contribute to more than 90% of all views on Facebook and Instagram' and excludes private conversations on Messenger and Instagram Direct; it is a sampling estimate, not a census, and Meta cautions that small shifts near the reporting floor may not be statistically meaningful. Content-actioned counts, meanwhile, exclude anything Meta blocks before it is ever posted, such as spam it stops from being created. Neither figure is broken out for video content, Reels, or any single format, so a claim that ties either number specifically to Reels moderation goes beyond what Meta's reporting states.

What to try next

Anyone citing Meta's enforcement data should specify which metric is in play and should not describe a rising content-actioned count as evidence that violations themselves increased, since Meta's own documentation says the two can move independently. This is an editorial caution, not a Meta rule: read prevalence as the platform's best estimate of what people actually saw, and content actioned as a measure of Meta's own enforcement activity.

• Is the figure being cited content actioned, a volume count, or prevalence, a sampled view-share estimate?
• Does the claim specify Facebook, Instagram, or both, and does the underlying report even separate the two?
• Is there a video- or Reels-specific breakout in the source, or is the number being extrapolated from an all-content figure?

Meta's own distinction between enforcement volume and violation prevalence is the most useful thing its transparency reporting offers, and it disappears the moment the two get used interchangeably.