• by Colin Kirkland , Yesterday

According to the Financial Times, 50% of content and ad reviewing across Meta's family of apps is currently being handled by LLMs.

By 2027, the company is planning to increase this number to over 90%, or almost all of the content and ad reviewing tasks.

In March, Meta rolled out its Meta AI assistant to help social-media users with simple support issues including password updates and profile settings, while also announcing plans to rely more heavily on AI to moderate content violations, including scams and illicit drug sales.

At the time, Meta stated that human experts would still be in charge of designing, training, overseeing and evaluating its AI systems, specifically with regard to “the most complex, high-impact decisions” such as “appeals of account disablement or reports to law enforcement.”

The tech giant has informed the Financial Times that since March, its tests have shown that on average, LLMs make 13% fewer mistakes than humans when moderating harmful content, ultimately uncovering 10% more violations.

In addition, Meta previously reported that its AI tools drove down views of ads with scams and other violations by 7%, offering potentially promising results and protections for brands.

“The point of this work is to improve our enforcement efforts, and we're deploying these more advanced AI systems once we're sure they're consistently performing better than our current methods of content enforcement,” a Meta spokesperson said.

However, replacing human review with AI could leave room for other issues, including a lack of reliable oversight.

For example, this past spring, Meta was sued after Kenyan workers employed by a Meta subcontractor reported seeing users of Meta's Smart Glasses “going to the toilet, or getting undressed” in intimate recordings.

Furthermore, AI agents have been known to act irrationally. In March, an AI agent went rogue at Meta, offering an engineer advice that led to the disclosure of sensitive company and user data to employees who did not have permission to access it. And earlier this month, 20,000 Instagram accounts were broken into by hackers who gained access through Meta’s AI support bot.

Advertisers may also be wary of a more complete shift toward automated ad reviewing, as Meta continues to face lawsuits over the alleged influx of scam ads on its social media apps.

In 2025, Reuters reported that Meta was secretly forecasting 10 percent of its annual revenue -- about $16 billion -- from scam ads.