Meta Moderation Shows Trade-Off Between Accuracy and Safety
Meta has released updated data on its content enforcement and account authenticity as part of its Community Standards Enforcement Report, detailing how its move away from traditional fact-checking toward a Community Notes model is playing out across Facebook and Instagram. This shift, first announced in early 2025, reflects a significant realignment of content moderation practices on the company’s platforms.
At the core of Meta’s latest reporting is a marked drop in incorrect removals of content. According to the company, less than 0.1 % of globally posted items were removed in error during the third quarter of 2025. Meta states that enforcement precision – which measures the share of correct takedowns versus mistakes – exceeded 90 % on Facebook and 87 % on Instagram. This, the company says, represents progress toward reducing complaints about wrongful takedowns in its automated systems.
These figures come amid Meta’s broader overhaul of moderation strategy, which involved ending its third-party fact-checking programme in the United States and implementing a community-led annotation system similar to the Community Notes used on X (formerly Twitter). Proponents within Meta argue this approach empowers users to add context to content without the platform itself acting as a unilateral arbiter of truth.
However, the latest report also reveals trade-offs. Meta’s proactive identification and removal of harmful content has declined in several categories, including bullying, harassment and hateful conduct, suggesting that some problematic material may remain visible to users more often than before. Meta attributes some increases in reported prevalence to changes in reviewer training and measurement methods, though critics suggest the outcome also aligns with a moderation strategy that prioritises visibility over removal of questionable posts.
The company also acknowledged that approximately 4 % of its more than 3 billion monthly active users are estimated to be fake accounts – a figure that translates to over 100 million profiles across its services. While Meta maintains that this aligns with industry norms, the sheer scale of these accounts raises ongoing questions about the platform’s ability to manage inauthentic behaviour effectively.
