A peer-reviewed study reports strong internal performance for an AI model distinguishing laryngeal lesions, but external validation is still needed.