The exact end-of-session prompt that forces Claude to hunt its own AI tells and rebuild your voice from the ground up.
What's inside
by Divjot Sahni · EZYE Consulting · ezye.com.au
The tells that give Claude away
The rhythm is too even — every sentence lands the same length and weight
Transitions are too clean, connective tissue that no real person writes
Tidy wrap-up lines that summarise what you just said for no reason
A handful of giveaway phrases that appear in almost every AI output
Why 'make it sound more human' fails
Asking Claude to sound human just adds more polish, not less
The model can't self-diagnose while it's still writing in its own voice
You have to force it to audit the output as a critic, then rebuild
Here's the mechanism that actually breaks the pattern...
Free access — enter your email below