Interviews with 17 UK young adults experiencing eating disorders revealed three distinct patterns in how TikTok's recommendation algorithm entraps vulnerable users in harmful content cycles. The platform's For You Page creates cascading exposure where viewing a single problematic video triggers an avalanche of similar content, exploits users' compulsive viewing behaviors despite recognized harm, and paradoxically converts recovery-focused content consumption into pro-eating disorder recommendations. This research illuminates a critical blind spot in social media platform design where algorithmic engagement optimization directly conflicts with user wellbeing. The findings suggest TikTok's recommendation engine lacks contextual awareness to distinguish between healthy recovery content and harmful pro-eating disorder material, treating both as similar engagement signals. For the estimated 9% of Americans who will experience an eating disorder, this represents a significant digital health hazard during crucial recovery periods. The study's limitation to self-reported experiences and UK participants suggests broader international research is needed. However, the consistent patterns across participants indicate TikTok's algorithm may be systematically undermining eating disorder recovery efforts, potentially requiring regulatory intervention or fundamental algorithmic redesign to protect vulnerable populations from harmful content amplification.