The feedback was that the rework felt like AI, not like her. So we built a fingerprint of her real voice across her own emails, texts, and pages, then wrote the rebuild to match it on every axis. Every figure below is computed by the same pipeline over her real assets, channel-aware and re-runnable.
Register, by channel
How often she writes it's, not it is
Contraction rate is the single signal that drove the AI-slop verdict. Her most intimate register lives in the P.S. and the text; her subject lines stay crisp. The rebuild matches her channel by channel rather than running one flat rate everywhere.
Readability
The phone-skim band
Flesch reading ease, with a grade level around . The plain-spoken band a busy parent reads on a phone. The rebuild sits in the same band, never drifting into corporate register.
Rhythm, the swing
Short punch, long breath, fragment
1 wordmedian85 words
of her sentences are five words or fewer; run past twenty. That swing is the texture. Coefficient of variation . The milestone-1 copy ran uniformly long and flattened it.
Emoji, her palette
Sparing, deliberate, placed
emoji per 1,000 characters across glyphs. The number under each glyph is how often it leads a line. The clock leads a real deadline; the sparkle and heart land on warm beats. The rebuild uses her exact palette at her exact rates.
Posture, the mood mix
She tells, rarely commands, rarely asks
She mostly states something true and lets it land. The blend of statement, command, and question is her persuasion posture, and the rebuild holds the same ratio.
Vocabulary, the spine
Pronoun-heavy, you-forward
Her top function words per 1,000 tokens. The spine of a personal voice is how often it says you and your. The size of each word here is its frequency. This is the skeleton every sentence hangs on.
Emotional arc, per campaign
The designed climb toward the close
Urgency Warmthper 100 words, across the send sequence
Each campaign warms and tightens on a deliberate curve. Urgency rises toward the deadline while warmth holds steady, so the close feels earned, not pushy. Tap a campaign to see its measured arc.
The whole shape
Her voice on six axes
The normalized voice shape across six measured axes: contraction, rhythm swing, lexical diversity, readability, command-and-question rate, and emoji density. This single overlay is what a rewrite is held against, at a glance.
Neighborhoods of her voice
What she keeps coming back to
Schematic stand-in for the live 3D embedding cloud
Her topic model groups her copy into four recurring neighborhoods: the member event, the little moments, the offer and pricing, the archive and gallery. Word size is its weight in that theme. This is a static schematic of the real topic clusters; the full interactive 3D embedding cloud renders once the embedding pipeline is back online.
Every figure is computed by the same channel-aware pipeline over of her real assets and is re-runnable from source. Nothing here is a vibe.