Forty-seven million enforcement actions is not a decorative statistic
Scribd finally put some hard numbers on a problem most document platforms prefer to discuss in foggy corporate language. Its first transparency report, covering January through June 2026, says the company took more than 47 million moderation actions across Scribd and Slideshare. That is a hulking workload, and it matters because this business lives or dies on whether users believe the documents they find are useful, lawful, and not riddled with junk.
The raw scale is what jumps out. Scribd says its two covered platforms hold more than 400 million pieces of content, serve 250 million monthly users, and operate across 50 languages. Another 84 million user-uploaded items were published in just the first half of 2026. Anybody who thinks this can be managed by a few moderators squinting at a queue has not paid attention to the internet for the last decade.
Automated systems blocked 42 million uploads before they appeared, then removed more than 5 million items that had already been published. That second figure includes older material, which is worth remembering. We do not know from this report exactly how much of the removal pile came from fresh uploads versus legacy content. Still, the direction is plain: Scribd is leaning hard on machine-assisted gatekeeping because there is no credible alternative at this volume.
The copyright bill is the part that deserves real attention
Of the roughly 5 million published items removed, about 2.9 million were tied to alleged copyright infringement. Scribd’s BookID system identified 2.76 million of those removals automatically, or about 95%. It also received 48,569 DMCA notices and removed 145,248 pieces of content in response.
That is a useful distinction. External complaints still matter, but Scribd is not merely waiting for rights holders to come knocking. Its proprietary system is doing most of the heavy lifting. From a sector-watch perspective, that is encouraging. Copyright exposure can turn a platform with a big catalog into a very expensive headache in a hurry. A functioning detection system does not erase legal risk, naturally. Nothing involving user uploads ever gets that clean. But catching suspect content at upload is cheaper and less damaging than untangling it after distribution.
There is an uncomfortable tradeoff tucked inside those figures, too. Automated enforcement can be fast, but fast systems can be blunt systems. Scribd reviewed 3,539 appeals and upheld 3,296 removal decisions, roughly 93%. That suggests the company’s decisions usually held up when challenged. It does not, however, tell us how many users gave up rather than appeal, how long reviews took, or how often a legitimate upload was initially blocked. Those are not accusations; they are the missing operational details that future reports should gradually fill in.
A baseline is useful only if management keeps showing up
The company deserves credit for publishing a report at all. This is its first in nearly 20 years, and the timing is not accidental. Platforms carrying documents, presentations, images, and material in multiple formats are under pressure to explain how they police abuse without choking off legitimate sharing. Scribd says the average document runs 24 pages. That is a different moderation animal from screening a short post or a single image.
Community reporting remains part of the machinery. More than 48,000 users filed 108,638 reports involving 52,359 pieces of content. Meanwhile, Scribd terminated 15,355 accounts under its repeat-infringer policy or for other copyright-related reasons. Those figures show that enforcement is not solely an algorithmic conveyor belt. Good. Community signals and human review are messy, but a platform that ignores either one eventually gets blindsided.
- Automation blocked the overwhelming majority of questionable uploads before publication.
- Copyright accounted for more than half of published-content removals.
- Appeal outcomes largely supported the original enforcement calls.
- The data covers Scribd and Slideshare, not Everand or Fable.
Do not stretch this report beyond what it says
There is one caveat investors and observers should keep taped to the monitor: this is only a half-year snapshot, and it excludes two other Scribd, Inc. products, Everand and Fable. It would be sloppy to treat these numbers as a complete readout on the parent company’s trust-and-safety posture. It would also be sloppy to annualize every figure and call it a trend. One reporting period is a baseline, not a verdict.
Scribd says it plans another report in 2027 and intends to expand its team, refine policies, improve classification, and add technology. Fine words are cheap; consistency is the test. If the next report breaks out error rates, review timing, repeat-offender patterns, and changes in copyright claims, it will be far more valuable. For now, this is a credible first look at a platform doing an enormous amount of unglamorous cleanup. Frankly, that work is not glamorous because it is essential.