AI-based moderation systems struggle with nuance like sarcasm and slang, often disproportionately flagging marginalized groups. Gilbert, research director at Cornell's Citizens and Technology Lab, notes that vulnerable populations face the highest moderation rates due to false positives from counter-speech and language reclamation. "False positives are an equity issue," she says, silencing already marginalized communities.

AI also undermines community self-moderation. On Reddit, if AI removes hateful content before human moderators see it, they lose the ability to assess whether bans are warranted. Reddit is testing Rules Hub, a tool suite letting human mods choose automated enforcement rules, preview actions, and review logs—eventually replacing the keyword-based Automod.

Moderators report a spike in rule-breaking content driven by generative AI. While companies seek better automated solutions, reducing human input is a step backward. Low-effort AI-generated content demands stronger approaches combining machine-scale detection with human judgment. Content moderation cannot succeed without human expertise at the forefront. Advance Publications, owner of Condé Nast, is Reddit's largest shareholder.