Your AI Writes Like Everyone Else

You are using AI to write, and it sounds like AI. Not bad — worse. It sounds like every other person using the same model with no instructions. Flat, safe, interchangeable. You could swap your output with a stranger's and neither of you would notice.

This is not the model's fault. It is yours. The model writes generic because you gave it nothing specific. Three fixes.

1. Feed it your voice before the task

Here are three paragraphs I wrote recently. Read them for voice only — not topic, not structure. Match this tone, sentence length, and attitude in everything you write for me from now on.

[paste your writing]

Most people skip this because it feels redundant. It is the single highest-leverage prompt you will ever write. The model is very good at mimicking style — the problem is that nobody gives it a style to mimic. Without a sample, it defaults to the median of its training data: corporate, cautious, and stuffed with transitions nobody reads.

Three paragraphs is enough. An email you wrote, a Slack message, a paragraph from a blog post. The model does not need a novel — it needs signal.

2. Ban the words it loves too much

Never use these words: delve, leverage, crucial, seamless, robust, elevate, streamline, cutting-edge, game-changer, empower. If a sentence needs one of them, rewrite the sentence.

These are the model's comfort words. They appear in every AI-generated text because they appeared in millions of training examples. Banning them does not make the writing good — it makes it yours, because your natural vocabulary does not include "delve into the nuances."

Add your own banned list. Read your last three AI outputs, highlight the words that made you cringe, and put them here. The list will grow. That is the point.

3. Ask for the rough draft, not the finished piece

Write this as a rough first draft. No transitions between sections. No introduction that summarises what follows. No hedging phrases like "it is worth noting" or "one might argue." Just the raw ideas, one after another.

The model's default is to write a finished essay. Finished essays from AI all sound the same because the finishing — the transitions, the hedging, the summary paragraphs — is where the generic lives. The ideas might be good. The wrapping is not.

Ask for raw. You do the wrapping. That is where your voice comes in: not in what you say, but in how you connect it. The model gives you the bricks. You build the wall.

The default is not neutral — it is generic

When you open a blank prompt and type "write me a blog post about X," the model does not choose a voice. It averages. And the average of a million writers is no writer at all.

Your voice exists. The model can match it. But it will never go looking for it — you have to bring it in. Three paragraphs of your own writing, a list of banned words, and a request for raw thinking instead of polished nothing. That is the fix. It costs one setup and changes every output after it.

Want the calm version of AI news like this, once a week? Subscribe to the Sharp AI Hub newsletter →