Don't outsource your personality to ChatGPT

I can usually tell when a brand has handed its personality to a model. The prose is polite, symmetrical, and strangely unable to dislike anything. Everything is exciting. Nothing is specific. It is the literary equivalent of a hotel lobby that smells of nothing in particular and therefore smells of everywhere.

Automation is useful. Authenticity is harder to automate. When every channel converges on frictionless enthusiasm, customers cannot tell you from the three competitors who made the same bargain. Voice dilution rarely arrives as a decision. It arrives as a series of convenient edits.

Bad AI habits vs good AI habits

**Bad SaaS brand:** homepage paragraphs that could belong to any B2B tool, hedged enthusiasm, and a blog that never names a trade-off. Readers feel nothing because nothing was risked. **Good product brand:** a short list of what you will not build, written in human language, then AI used only to tighten structure around that spine.

**Bad consumer brand:** captions that begin with “Excited to announce” and end with five hashtags the model suggested. **Good outdoor or hospitality brand:** one memorable constraint (weather, neighbourhood, repair) that a model cannot invent without your field notes.

What to do this week

1. Write a one-page voice brief: three words you use, three you ban, one belief you will defend in public.
2. Collect ten sentences customers actually said. Make them mandatory source material before any model draft.
3. For the next five pieces, generate a draft only after you write the ugly first paragraph yourself.
4. Add a review gate: if the piece has no specific noun, number, or refusal, it does not ship.
5. Audit last month’s content. Flag anything a competitor could publish unchanged. Rewrite those first.

For practical writing research on how people scan and decide, see Nielsen Norman Group. Related: AI can write your content; it can’t replace your perspective.