Users trust AI as much as humans for flagging problematic content