This paper argues for machine-readable science and demonstrates its utility with a journal-scale comparison of AI and human peer review.