Mistake L2 · Context engineering informational

Don't Let It Flatten Your Voice

What it is

An awareness habit: ChatGPT edits toward a smooth, generic mean, so instruct it to fix errors while preserving your voice — your rhythm, word choices, and quirks — rather than sanding them off.

edit, don't homogeniseYour draft, with voiceConstrain the editFixed but still yours
Correct the errors, keep the fingerprints that make it yours.

Why it works

The model's default edit removes not just mistakes but personality: distinctive phrasing gets 'clarified', varied rhythm gets evened out, and the result reads competent and anonymous. Telling it to correct only genuine problems and leave your stylistic fingerprints keeps writing that sounds like a person instead of an assistant.

When to use it

Editing anything where voice is the value — personal essays, opinion pieces, brand copy, your own distinctive writing.

When not to use it

Purely functional text — instructions, technical docs, forms — where a neutral, standard voice is exactly what you want.

Prompt

Edit this for genuine errors only — grammar, clarity, obvious clunk. Preserve my voice: keep my word choices, sentence rhythm, and any deliberate quirks. If something reads oddly but is clearly intentional, leave it. Text: <paste>.

Example

Instead of smoothing your punchy fragments into full sentences, the edit fixes two real typos and leaves the staccato rhythm that made the piece feel alive.

Common mistakes

  • Accepting edits that trade your voice for generic polish.
  • Not flagging deliberate stylistic choices, so they get 'corrected'.
  • Over-editing personal writing until it sounds like every other AI draft.

Related