The great flattening

There is a version of editing in which every interesting sentence is gradually replaced by one nobody can object to. The strange detail becomes “a key challenge”. The disagreement becomes “an opportunity to align”. The customer’s actual problem, which was awkward and specific and might have made the reader care, becomes a journey.

This happened before the machines. Committees managed it perfectly well with tracked changes and biscuits. Generative AI makes the movement towards an acceptable average unusually quick. Ask for something more professional, then warmer, then punchier, then more concise, and watch the person who had something to say disappear behind the improvements.

The test is not whether the result sounds impressive in isolation. Read it beside three other pieces on the same subject. Can we tell who wrote it? Does it contain a particular observation, a decision or a conviction? Would anything be lost if we replaced it with another plausible version?

Distinctiveness is not a licence to be confusing. We still owe the reader an intelligible argument. But clarity should reveal the particular thought more precisely, rather than sand it down until it fits every other organisation’s announcement.

For the writer, this means protecting a few things deliberately: the expert’s actual explanation, the inconvenient fact, the phrase that sounds like somebody saying it aloud. It means resisting the suggestion that every sentence needs the approved finish. Some sentences need to remain slightly awkward because the thing they describe is awkward.

We can use a machine and keep a voice. It requires us to know what belongs to us before we invite the machine to improve it.