The One Prompt That Fixed My Writing (After a Year of Bad Ones)
Forget 'act as a copywriter.' The prompt that worked told the model what NOT to do.
I spent a year writing prompts that asked chatbots to 'improve' my drafts. They always came back polished and hollow, full of words like 'leverage' and sentences that could belong to anyone.
The breakthrough was negative. Instead of describing the voice I wanted, I listed the voice I hated: no hype words, no three-em-dash fake insights, no 'in today's world' openings, no paragraphs that all start with 'moreover.' The model finally had edges to work inside.
The second change was constraint. I told it the draft must stay under a word count and keep my original opening line untouched. Forcing it to respect a frame stopped the generic rewrite and kept my actual thinking in the piece.
The result reads like me, just tighter. Not because the model is smarter, but because I stopped asking it to be a better writer than me and started asking it to be a stricter editor than me.
If your AI writing sounds like everyone's AI writing, the problem is your prompt's missing the 'don'ts.' Tell it what to avoid and watch the personality come back.