Reddit uses LLMs to cut spam exposure 20% and block 23 million views daily
Reddit announced upgraded AI defenses that deploy large language models to detect coordinated fake behavior, blocking about 23 million spam views each day and catching roughly 25 000 new spam posts. The rollout has already lowered user spam exposure by 20% and revoked nearly 2 million inauthentic votes daily.