Readers have become good at spotting AI-written text. It tends to be smooth but empty: lots of “in today's fast-paced world”, neat lists of three and conclusions that restate everything. The fix isn't to stop using AI; it's to change how you use it.
Give it your raw material
Generic input produces generic output. Instead of “write a post about remote work”, give the AI your actual opinions, a story from your week and the specific point you want to make. The AI is excellent at shaping ideas, but the ideas should be yours.
Show it how you write
Paste two or three samples of your own writing and ask the AI to describe your style: sentence length, tone, words you use often, how you open and close. Then ask it to write in that style. Saving this style description and reusing it is one of the most effective habits you can build.
Ban the clichés explicitly
Tell the AI what to avoid. A short list works well: “Don't use the words delve, unlock, elevate or landscape. Don't start with a rhetorical question. Don't end with a summary paragraph.” Models follow concrete instructions far better than “sound natural”.
Ask for specifics
Add instructions such as “use one concrete example”, “include a number” or “mention a real situation the reader will recognise”. Specific details are what make writing feel human.
Edit in your own voice
Treat the AI's version as a first draft. Read it aloud, cut anything you wouldn't say in conversation, and vary the rhythm: a short sentence after a long one adds energy. Five minutes of editing makes a bigger difference than any prompt.
A prompt you can reuse
Rewrite this in my voice: [paste style description].
Keep it specific and conversational. Vary sentence length.
Avoid: [your banned words]. Keep all facts.
Text: [paste draft]