Zarema Khalilova · zmoki.xyz
Voice fingerprint analysis
Your voice is unique. AI rewrites it in its own handwriting. This analysis preserves yours.
I measure what makes your writing yours, so your voice survives in every post you publish with AI.
Analysis plan
1
Baseline
your idiosyncrasies, measured from texts in your original voice
2
Score
every Substack and LinkedIn post against your baseline, showing how far AI pulled each toward its own voice
3
Improvements
how each published post should sound, and what to change to get there
4
Guidelines
how to keep publishing with AI and keep sounding like you
Methodology
Voice lives in structure, deeper than word choice.
So I measure 20 discourse-level features built on two published studies.
StoryScope · COLM 2026. Found AI's idiosyncrasies survive style editing because they live at the discourse level.
NarraBench · EACL 2026. The taxonomy of narrative understanding the features are drawn from.
10 core features · where AI flattens any writer
how a post opens
concrete specifics
how it ends
emotion handling
hedging
transitions
sentence rhythm
10 fingerprints · your idiosyncrasies, patterns only you produce
recurring themes
signature phrasings
real people & numbers
how you address readers
your quirks
What you get
Visual report
Your baseline overview. How far AI pulled toward its own style. All posts ranked, all gaps mapped.
post 14
post 09
post 21
post 03
post 17
voice flattened
drifting
fully you
Improvement plan
For the posts already published. Flagged passages quoted, and exactly how to rewrite each one to bring your voice back.
Brief for future posts
Your editor's guide for writing with AI, so every new post keeps sounding like you.