A sock-puppet audit of 323 TikTok accounts found the recommendation system supplied 11.5 percent more in-party content to Republican-trained bots than to Democratic-trained bots between April and November 2024, independent of initial user signals or state location. Researchers at NYU Abu Dhabi created weekly cohorts of Android bots aged 22-24, trained on up to 400 partisan videos for one minute each, then recorded the first 10 seconds of 280,000 For You recommendations; transcripts from 40,264 videos were labeled by an ensemble of three language models validated by political science students (PsyPost, 2025; Rahwan et al., Nature). Neutral control bots in Georgia and location-masked accounts in New York and Texas produced the same asymmetry. The same lab previously isolated algorithmic effects on YouTube using comparable methods, noting persistent difficulty separating platform push from user choice (Rahwan, 2023). The TikTok results parallel internal 2020 platform measurements reported by Facebook researchers showing differential partisan reach after engagement optimization. Weekly device resets eliminated cross-week tracking, confirming the observed skew originated in the automated feed rather than persistent user profiles.