Reddit uses LLMs to fight LLM spam

Reddit developed tools using large language models to combat spam, much of which is also created with LLMs. The platform blocks 23 million spam views per day and catches about 25,000 new spam posts and comments daily. Reddit reduced users' exposure to spam by 20% from January to March compared to the prior three months.