Research: 23 of the top 30 search results for the numbers 11, 12, 13 on Facebook were groups targeting children of those ages, including with sexual references
& with rising disbelief found it impossible to get them taken down. This is my story. https://www.wired.com/... Quinns / @quinns108 : “Meanwhile, as Facebook's recommendation engines function like a seamless Uber for abusers, the safety side functions like the DMV circa 1990: manual data entry, inaction as default.” https://twitter.com/...
Context & Ripple Effects
This research lands on a pattern Wired's source documents in detail: Facebook's search and recommendation machinery reliably connects abusers with child-targeting groups, while the reporting-and-removal side stalls — the researcher describes being unable to get the groups taken down at all. It is not a first failure. Facebook already apologized for its search box suggesting sexual videos and child abuse content back in 2018, meaning the same surface has now failed across two product generations.
The finding also sits inside a documented pipeline: algorithms that help bad actors find each other were previously shown with closed CBP and extremist Facebook Groups, and the same recommendation logic later surfaced underage sex content on Instagram and Facebook Groups even after Meta stood up a child-safety task force.
First-order effects
- Children searching or being searched around ages 11–13 are directly exposed: 23 of the top 30 results for those numbers were groups targeting kids of exactly those ages, some with sexual references.
- Meta's trust-and-safety operation is shown unable to act on researcher reports of child-targeting groups — manual review is the bottleneck while automated recommendations do the connecting.
Second-order effects
- The enforcement vacuum feeds extra-platform responses: vigilante groups staging live-streamed sting operations exist precisely because official takedown channels fail, moving child-abuse policing from Meta to amateurs with no due process.
- Regulators and journalists now have a repeatable test — run a search, count the results — which turns each new investigation (Pinterest's kid-themed sex boards among them) into comparable evidence against multiple platforms rather than one-off scandals.
Third-order effects
- If the asymmetry holds — recommendation engines scaling instantly, human review stuck as default-inaction — platforms will face legally mandated design duties on search and ranking for minors, not just content-removal obligations.
- The recurring findings suggest child safety cannot be fixed by task forces bolted onto growth-optimized systems; expect structural separation between recommendation surfaces and minor-related queries to become the demanded fix.
The trend: Platform recommendation systems are outpacing their own safety apparatus so consistently that child-surface failures are becoming a regulatory-design problem rather than a moderation problem.