Ethics & Responsible Use
Using AI without losing your own voice
AI writing has a flavour, and readers can taste it. Here is how to use these tools heavily and still sound unmistakably like yourself.
Updated 8/14/2026
There is a specific texture to AI-assisted writing that people have learned to detect. Not errors — errors would be easier. It is smoothness. Every sentence a similar length. Every paragraph a tidy three-part structure. Transitions that are technically correct and emotionally inert.
Readers cannot always name it. They can always feel it. And once they feel it, they stop trusting that anyone is home.
Here is how to use these tools constantly and not end up there.
Understand what is actually happening
A model produces the statistically most reasonable continuation. Your voice, by definition, is where you deviate from reasonable. The tool is not stealing your voice maliciously; it is averaging it out, one sensible sentence at a time.
That reframing matters, because it tells you exactly where to intervene: at the places you would have been idiosyncratic.
Draft first, then assist
The single highest-impact rule. If the model writes the first draft, you spend the rest of your time editing someone else's structure. Your best ideas — the odd angle, the weird example, the joke — arrive during first-draft friction, and delegating that friction deletes them.
Write badly, quickly, yourself. Then bring in help.
Use it for the parts that are not voice
Some writing work is genuinely voiceless, and outsourcing it costs you nothing:
- Restructuring an argument you have already made
- Finding where you repeated yourself
- Tightening a paragraph you have stared at too long
- Translating jargon into plain language
- Being an aggressive, tireless first reader who tells you which section is weak
That last one is the killer application and it is chronically underused. "Which paragraph would a bored reader skip?" is worth more than any generated draft.
Ban the tells
Keep a standing list in your prompts of words and moves you do not use. Everyone's list is different; the point is having one. Ours includes phrases like "delve into," "in today's fast-paced world," "it's not just X, it's Y," and the compulsive tricolon.
Also ban the structural tells: an intro that announces what the piece will cover, a conclusion that recaps it, and section lengths that are all suspiciously equal.
Keep the friction that makes you *you*
Your specific memories, your specific opinions, your specific jokes — a model cannot produce these because it has never had them. Write those parts unaided, always. If you find every personal anecdote in your work has been smoothed into a general observation, you have gone too far.
The read-aloud test
Read the finished thing out loud. Not silently — actually aloud. Anywhere your mouth trips, or where you hear a voice that is not yours, rewrite by hand. It takes four minutes and it catches nearly everything.
The line that matters ethically
Using a tool to write better is fine. Using it to appear to be someone you are not is not. The distinction is not about how much AI was involved; it is about whether the finished thing represents your actual thinking and whether you would stand behind every claim in it.
If you would not defend a sentence in conversation, it should not be in your writing, no matter who generated it.
The point
The goal is not less AI. It is that the parts of your work that only you could have made stay yours. Everything else is admin.
Use the tools. Keep the voice. Both, ticked.
Keep going
Build something with the prompt generator, decode the jargon in the glossary, or compare the tools on our platform deep-dives.