Module 1 · Section 5 of 5
Exercise: Find the Cringe
Here’s an uncomfortable truth: Your AI writes the right words, but sounds nothing like your business.
Most people give Claude or ChatGPT examples of what they want to sound like, but never tell it what to actively avoid. The result? Content that’s technically correct but tonally wrong.
The missing piece is negative training.
When I developed the Signal Over Noise voice, the breakthrough didn’t come from more positive examples. It came from systematically documenting what I never wanted to sound like.
Example from my forbidden phrases list:
- “Navigate the complex landscape of AI implementation”
- “Here’s what actually works in AI implementation”
The first sounds like every other AI consultant. The second sounds like me.
Your task (5 minutes):
Find 3 pieces of AI-generated content that make you cringe. Could be LinkedIn posts, newsletters, website copy — anything that screams “this was written by AI.”
For each piece, write down:
- What specific phrases feel obviously artificial
- What tone mistakes damage credibility
- How you would never say the same thing
Save these examples. In the next module, we’ll use them to start building your “never list” — the secret ingredient that transforms generic AI output into something with actual voice.
Most people want to skip straight to the AI prompts. But without understanding what you actively reject, you’re just creating slightly more personalised generic content.
The magic happens when you train AI on both what you do sound like AND what you never sound like.