“Concise”, “professional” and “engaging” do not survive the trip

A word that means something different at both ends

Ask for a concise summary and you will get one, in the sense that it is shorter than the source. Whether it matches the length you had in mind is a coin toss, because concise isn’t a measurement — it is a comparison against an expectation you never stated.

The same goes for professional, which resolves towards a bland corporate register, and engaging, which resolves towards short sentences and a rhetorical question near the top. Each word does something. What it does is the average of how it is generally used, and the average is rarely what you wanted.

What makes this hard to spot is that the output is never wrong exactly. It is a defensible reading of the adjective you supplied, which means the correction you type next is another adjective, and the second one goes the same way as the first.

Replace the adjective with a decision

Every quality word can be unpacked into choices someone actually made. Concise usually means one of: under a certain length, no preamble, one idea per paragraph, or no restating of the question. Those are four different instructions and they produce four different outputs.

Professional often means no exclamation marks, no contractions, no metaphors, and the conclusion stated in the first line. Engaging often means a concrete example early and no abstract nouns as sentence subjects. When you write the decisions instead of the label, the result stops drifting.

The exercise is worth doing once for the words you use most. It takes ten minutes, it produces two or three lines you will paste for years, and it usually reveals that two of your favourite adjectives were pulling in opposite directions the whole time.

Reference beats description

Where the quality is genuinely hard to articulate — and voice usually is — the shortcut is to supply an example rather than a definition. Two or three paragraphs of writing that already has the property you want will constrain the output more tightly than a page of careful adjectives.

It works because the sample carries the small decisions that nobody thinks to describe: sentence length, whether the register admits a joke, how much is assumed of the reader, whether the first person appears. You could not have listed those. You didn’t have to.

The trap is that a sample constrains content as well as manner. If the example is about a different subject, say so explicitly and say which aspects to take from it, or you will get your subject bent towards the example’s topic in ways that are hard to unpick.

Some vagueness is load-bearing

It is possible to over-specify. Pin every dimension of tone and the output comes back stiff, self-conscious, and obviously assembled to satisfy a list. Prose written to eleven simultaneous constraints reads exactly like prose written to eleven simultaneous constraints.

The distinction to hold is between qualities that matter to a reader and qualities you merely have an opinion about. Length matters. Whether the conclusion comes first matters. Whether a particular connective appears in the third paragraph doesn’t, and pinning it costs you fluency for nothing.

A rough guide: specify the two or three properties whose absence you would actually complain about, supply a sample for the rest, and let everything else vary. That produces output you can edit rather than output you have to rescue.

Judging the result needs the same discipline

The adjective problem does not stop when the output arrives. Reading a draft and thinking it feels off is the same failure in the other direction — a precise reaction expressed as a word that carries no information back into the next attempt.

Push yourself to name what is wrong as a decision. It buried the point. It is a hundred words too long. It explains something the reader knows. Each of those maps directly onto a correction, whereas make it punchier maps onto a shrug and another round.

Do this for a few weeks and the vocabulary you develop is useful well beyond the tool. It is, more or less, what editors mean when they say they can tell you why something is not working, which was always the harder half of the job.

There is a limit worth admitting. Some reactions genuinely resist articulation, and a draft can be wrong in a way you will only be able to name after rewriting it yourself. When that happens, rewriting is the faster route, and pushing the feeling through three more rounds of adjectives is how an afternoon disappears.