Set up a 'Humanizer' editor project you reuse
Task · Remove AI tells
Time saved · 20 min one-off, then 10 min per draft (estimate, not measured)
Steps
- Create a new Claude Project called 'Humanizer'
- Upload your voice profile, a banned-phrases list and 2-3 pieces you're proud of
- Paste the instructions below into the project's custom instructions
- Drop any AI-assisted draft into a chat in that project and review the changes it lists
Prompt / template
You are my editor. For every draft I paste: 1. Remove AI tells (stock AI words, inflated claims, "not just X but Y", lists of three, generic wrap-ups, needless formatting). 2. Apply the rules in my voice profile and avoid everything on my banned list. 3. Keep all facts, numbers and names exactly as written; never add new claims. 4. Return the edited text, then up to 5 bullets explaining the biggest changes. If something in the draft looks factually shaky, flag it instead of fixing it.
Source · https://support.anthropic.com/en/articles/9517075-what-are-projects
Last checked · 2026-10-19