I used to think of voice as something a writer either has or does not have. Now I think of it as a list of habits. Once I could list mine, I could keep them steady across posts and hand them to a tool. This post shows how I describe a voice.
Voice is a pattern, not a talent
Read three posts by the same writer and you start to predict the fourth. That prediction is the voice.
It comes from small choices repeated many times. How long the sentences run. Which words show up again and again. Whether the writer talks to the reader or about a subject. None of this is magic, and all of it can be written down.
The five things I look at
When I describe a voice, mine or anyone’s, I check five things.
Sentence length. Mine are short. Most are under fifteen words.
Point of view. I write as “I” and speak to “you”. Some writers use “we”. Some never appear in their own text.
Word choice. I pick the common word. I say “use”, not “utilise”. I say “help”, not “empower”.
Structure. I open with the problem, use five or six headed sections and close with one action. I explain that shape in how I structure a blog post people finish.
Attitude. I am direct and a little dry. I do not cheer, and I do not warn.
How I find my own habits
I cannot see my voice by thinking about it. I have to look at evidence.
I take three posts I like and read them slowly. I mark every sentence that sounds most like me. Then I ask what those lines have in common. The answer is usually simple. They are short, they make one claim and they do not hedge.
I also mark the lines I would cut today. Those show me the habits I want to drop.
Writing it down
A voice description should fit on one page. Mine has three parts.
The first part is the five points above, one line each. The second is a list of words I never use. The third is two sample paragraphs that show the voice working.
The samples matter most. A rule tells a writer what to do. A sample shows what done looks like.
Why this helps with AI
A tool cannot guess my voice from the word “casual”. It can match a pattern when I show one. That is the method in how I use AI to write without losing my voice.
That is why a system like Sastras reads existing posts before it writes. The posts are the description. The more consistent they are, the better the match. If my old posts are in five different styles, the tool learns a blur.
What I would do first
Take one post you are proud of and count the words in ten of its sentences. Write down the average. That one number tells you more about your voice than any adjective, and it is the easiest habit to keep steady.