How it works

Researchers presented trained reviewers with AI-generated and human-written novels and asked them to classify each. Detection accuracy was measured across reviewers.