Style Is the Last Fingerprint

One can recognise a friend from a single text message. Not from the news in it, but from the way it is written: the short sentences, the repeated word, the joke that only she would make. Every writer leaves a fingerprint, and it is not what they say. It is in how their sentences move.

I think about this often now, because so much of what we read is no longer written by anyone in particular. The machine writes beautifully. Too beautifully, sometimes. Its sentences are all the same length, all equally smooth, all equally safe. It never stumbles. It never repeats a word out of fondness. It never leaves a sentence half-finished because the thought ran ahead of it. Human writing does all of these things, and that unevenness is the fingerprint.

For my MPhil thesis this year, I asked sixty-nine readers to read six passages and say who wrote them — a person or a machine. Nearly half the time, they answered honestly: they could not tell. That result is usually reported as a warning, and it is one. But the other half of the result matters just as much. When readers did sense a difference, it was rarely the argument that gave it away. It was the style. The human passages had rhythm — a long sentence followed by a short one, a sudden pause, a word chosen for sound as much as sense. The machine passages had polish, and polish alone.

This is why I have come to distrust the current panic about detection. Universities and offices are buying software that promises to spot the machine, and the software keeps failing the wrong people — often the students whose English is already judged most harshly, whose careful, formal sentences look “machine-like” to a tool trained on somebody else’s idea of natural. A detector does not hear style. It counts it. And counting is not reading.

There is a better skill, and it is an old one: learning to hear a sentence. When you read something that feels too even — every paragraph the same shape, every transition perfectly oiled, every opinion stated with the same confident balance — slow down. Ask what is missing. Usually it is a lived detail: a specific place, a name, a memory that only the writer could have. Machines borrow the shape of experience; they cannot borrow the experience itself. A student who writes about her own street, her own teacher, her own failure in her own uneven words has given you something no generated paragraph can fake, however smooth it sounds.

I do not mean that we should become suspicious of every polished sentence. Some people write smoothly by nature, and some machine-edited work is honest work, openly done. The point is narrower: style is evidence, and we have stopped treating it as such. We outsourced judgement to software because software feels objective, and in doing so we stopped trusting the oldest instrument we have — the reader’s ear.

That ear can be trained, and it should be, in homes and classrooms alike. Read it probably was written by no one in particular. If it sounds like someone — with habits, hesitations, and a favourite word used once too often — believe that sound. It is the last fingerprint we have left, and it is still, just barely, human.