In recent months, a curious form of confidence has entered literary conversations. A sentence, a paragraph, sometimes an entire chapter is examined, and a verdict is delivered with striking precision: “78 per cent of this text is written by AI.” It sounds authoritative. It sounds measured. It sounds, above all, certain. But it is none of these things. There is, at present, no reliable method that can determine such a percentage with any degree of certainty. What is presented as measurement is, in reality, pattern recognition — an interpretation of rhythm, predictability, and linguistic regularity. These systems do not identify authorship; they identify statistical resemblance. And resemblance is not proof. A clear, structured, and reflective prose — especially one shaped by years of disciplined writing — may easily resemble what these systems classify as “AI-like”. In that sense, the accusation often reveals less about the text than about the limitations of the tool used to judge it.…