Reddit deploys LLMs to combat AI-generated spam

The company reports that its updated detection suite now blocks 23 million spam views daily, successfully intercepting approximately 25,000 new spam posts and comments every 24 hours. By leveraging LLMs to analyze coordinated patterns that previously bypassed legacy filters, Reddit claims a 20% reduction in user exposure to spam between January and March compared to the preceding quarter.

While this technological shift mirrors strategies at platforms like Meta and TikTok, experts maintain that algorithmic filtering remains incomplete without human oversight. The challenge lies in scaling these detection capabilities to address not only spam but also complex policy violations like hate speech, where the nuance of human moderation remains a critical safety layer.