Meta is replacing human content moderators with large language models (LLMs) to automate content review across India’s 22 official languages, aiming to cut costs and improve efficiency. The company has already automated about 50% of its human review requests this year and expects that figure to exceed 90% for some content types by the end of 2026, according to medianama.com.
The shift to AI moderation is part of Meta’s broader strategy to reduce its $5 billion annual content moderation spend, mostly on third-party contractors, and reallocate resources toward artificial intelligence development. Meta’s Q3 2025 Integrity Report highlighted that enhanced AI models require less human review for routine cases, allowing human moderators to focus on sensitive and complex enforcement. Tests showed LLMs outperforming both previous automation and human reviewers in detecting celebrity impersonation scams, per medianama.com.
India is Meta’s largest market by user base, spanning Facebook, Instagram, and WhatsApp, making effective moderation critical amid the country’s linguistic diversity. The challenge lies in whether AI can accurately police abuse across multiple languages, as current systems perform best in English. This move reflects a broader industry trend of increasing reliance on AI for content moderation, balancing scale with accuracy in diverse markets, according to medianama.com.
Meta’s transition to AI-driven moderation in India is expected to accelerate through 2026, with the company focusing human expertise on complex cases. The company’s 2025 content moderation expenditure and the planned automation ratio provide concrete markers of this shift, as detailed in Meta’s Q3 2025 Integrity Report and medianama.com.