A disclosed Claude-versus-Pangram experiment explores what an AI detection result can establish and what it cannot explain about authorship.