Tech Desk: Instagram users are spending over 10% more time on the platform than a year ago, a surge Meta attributes directly to its advanced AI recommendation systems. This isn’t just passive browsing; it’s a meticulously engineered experience designed to keep you scrolling, often leading to the phenomenon known as ‘doomscrolling’. Meta CEO Mark Zuckerberg and CFO Susan Li highlighted this increased engagement in their recent Q2 earnings call, revealing the profound impact of their algorithmic shift.

At the heart of this engagement boost lies a redesigned recommendation engine leveraging large language models (LLMs). This system deeply analyzes public Reels and Feed posts, dissecting their topics, tone, relevance, and quality. But it doesn’t stop there. The Meta AI algorithm meticulously tracks user behavior—from videos watched to completion, posts liked, shared, or saved, to frequently replayed Reels and account interactions. This comprehensive data allows the AI to predict precisely which content will maximize your time on the app.

While effective in boosting engagement, this powerful AI has drawn significant scrutiny. Meta is currently embroiled in several legal battles concerning the addictive nature of its platforms and their effects on young users, even setting aside $2.40 billion for related legal expenses. As Meta plans to extend a similar LLM-powered recommendation system to Facebook, the debate intensifies around the ethical implications of algorithms designed to optimize for screen time.