Prompt library
Define brand voice attributes with edges
Attributes with but-not edges and paired over-the-edge examples are what make voice guidance operational; bare adjective lists produce whatever the writer already writes. Testing the voice in hard moments (errors, incidents, price increases) defines it where brand trust is actually won or lost.
Last reviewed July 17, 2026
The prompt
Define our brand voice from the evidence below.
Evidence (copy we like, copy we hate, how we talk in person, competitor voices): {{evidence}}
Produce:
1. Four voice attributes maximum. Each gets: the word, the sentence-long definition of what it means HERE (because "confident" means something different at a bank and a skate brand), and crucially the "but not" edge: confident but not boastful, warm but not chummy. The edge is where voice lives.
2. Per attribute, a this-not-that pair: one sentence written in our voice and the same sentence just over the edge, so writers calibrate on examples, not adjectives.
3. The stress tests: how this voice sounds in the 4 hard moments (an error message, a price increase, a security incident, saying no to a feature request). Voices are defined by their worst days, not their homepage.
4. What our voice never does: 5 concrete bans harvested from the copy-we-hate evidence (specific constructions, punctuation habits, hedge words).
5. The one-paragraph writer's card: everything above compressed to something a freelancer reads in 60 seconds before writing.
If the evidence conflicts (we like copy that violates our own instincts), surface the conflict; do not average it away.Run in idaptOpens a new chat with the prompt prefilled. Nothing sends until you press send.
Fill in the variables
| Variable | What it is | Example |
|---|---|---|
| {{evidence}} | Copy samples you love and hate, plus context | [paste 5 pieces of copy you like, 3 you dislike, and your notes] |