Are 'AI Slop' Commenters on Reddit Actually Data Filtering Bots?

PresentSituation8736 · reddit · 2026-08-09

The author hypothesizes that Reddit accounts systematically posting 'AI slop' comments might actually be linked to automated data evaluation and filtering systems.

Experiments showed that made-up phrases posted on Reddit quickly appeared in the output of an AI model in an isolated session, indicating that external data seeps into model corpora very quickly. The author believes 'AI slop' comments from high-karma accounts might be identifying and filtering AI-generated text with specific statistical features to prevent training data contamination. The author also found that running AI-generated text through a translator before posting effectively evaded these detection systems.

Original post →

More from AGI Musings

AGI Musings channel →