AI writes publishable academic reviews

Frontier generative models can produce book reviews and critiques that academic authors judge comparable to human work, including substantive objections and conceptual engagement. That capability was demonstrated when an author asked OpenAI’s GPT‑6 Astra to write a 2,000‑word, journal‑style review of an Oxford University Press philosophy book and found it incisive and human‑grade. — If models reliably produce publishable scholarly criticism, journals, hiring, peer review and the authority of academic expertise will all face practical and ethical disruption.