How Trends Can Hide Automation
Automation blends into legitimate momentum when trend velocity is high, so we score participation quality, not just participation volume.
Trending topics create natural cover for scripted behavior. High baseline activity makes abnormal posting cadence harder to spot by eye.
We counter this by measuring interaction diversity: who someone replies to, how language adapts over time, and whether response depth changes with context.
Accounts that post constantly but never meaningfully adapt are elevated for review, even when they stay within surface-level content policy.
This approach reduces false negatives during major events without penalizing legitimate users who simply post often.
Trends are still valuable for discovery. The challenge is separating crowd energy from coordinated automation riding on top of that energy.