YouTube community guidelines report: 5M videos and 2M+ channels were removed between Oct. and Dec. 2019; 109K video removals were appealed and 23K reinstated
Of 109,000 removals appealed, YouTube reinstated 23,000 — In three months, YouTube received nearly 110,000 appeals from creators … Tweets: @stokel and @loudmouthjulia Tweets: Chris Stokel-Walker / @stokel : YouTube removed 963 videos a day in the last three months of 2019 for hate speech, 10,124 a day for child safety and 877 a day for violent extremism. But please, tell me how Ofcom can handle this https://twitter.com/... Julia Alexander / @loudmouthjulia : New: YouTube has a community guidelines report out. First time YT released numbers on creator appeals. 5M videos were removed last quarter. 110,000 removals were appealed. Approximately 23,000 reinstated. So YouTube tells creators to appeal, works less than 25 percent of time. https://twitter.com/...
Context & Ripple Effects
This is the fourth community guidelines enforcement report since YouTube began publishing them with its first disclosure covering Q4 2017, but it is the first to break out creator appeals — the piece of the system Susan Wojcicki promised to open up when she committed to more transparency and a better appeals process in early 2018.
The timing matters: hate-speech takedowns had already surged roughly fivefold after the June 2019 policy tightening (100K+ videos and 17K channels removed by September), so the new appeal data is the first public check on how often that expanded enforcement gets overturned.
First-order effects
- Creators finally get a number they never had: of 109,000 appealed removals, 23,000 were reinstated — roughly one in five — giving channels contesting takedowns a concrete success rate instead of a black box.
- The report also quantifies enforcement scale for regulators like Ofcom, which the coverage explicitly invokes: nearly 10,000 child-safety removals per day is a volume no national regulator could review itself.
Second-order effects
- With machine-flagged volume this large (the first report showed 6.7M of 8M removals were machine-flagged), appeals become the de facto human-review layer — pushing YouTube to staff and prioritize review queues where errors are most costly, like child safety versus borderline hate speech.
- Rival platforms now face pressure to publish comparable appeal-outcome data; once one leader discloses reinstatement rates, silence starts reading as concealment.
Third-order effects
- If the pattern holds, self-published enforcement reports evolve from PR into compliance infrastructure — platforms effectively setting their own audit standards before regulators do, with appeal rates serving as the metric policymakers reach for first.
- The structural endpoint is moderation built as flag-at-scale-plus-appeal-backstop, which concentrates power over what stays online in whoever operates both the classifiers and the appeal queue.
The trend: Platform transparency reporting is maturing from raw takedown counts toward outcome metrics like appeal reinstatement rates, pre-empting regulator-defined accountability standards.