⚡ Promptolis Original · Writing & Editing
👤 Ghostwriter Voice Matcher — Capture Anyone's Voice From 3 Samples
The structured voice-analysis system professional ghostwriters use to capture a subject's voice from 3 writing samples — covering lexicon, syntax, rhythm, rhetorical moves, and the 7 voice-signals that distinguish one writer from another with uncanny accuracy.
Why this is epic
Voice-matching is the highest-paying skill in professional ghostwriting ($50K-$500K+ per book). Most ghostwriters fail at it — their output reads as THEIR voice, not the subject's. This Original gives you the specific 7-signal analysis framework professional ghostwriters use: lexicon, syntax patterns, rhythm, rhetorical moves, register, fillers/tics, and thematic obsessions.
Produces a reusable voice-profile document you can reference while writing — so every paragraph you ghostwrite can be checked against the voice signature. This is how professionals maintain voice across 80,000-word manuscripts without it slipping.
Names the voice-matching tests (3 paragraph Turing test, lexicon overlap check, rhythm fingerprint) that catch voice-drift before the subject catches it. Based on professional ghostwriting workflows from writers who've ghosted NYT bestsellers + memoirs for executives, celebrities, and thought leaders.
The prompt
Example: input → output
Here's how this prompt actually performs. Real input below, real output from Claude Opus 4.
<subject>Farah Nasser — founder of a mid-sized fintech (B2B payments, 280 employees). Age 44, Pakistani-Canadian, based in Toronto. Known for direct communication style. I'm ghostwriting her first book on founder-led growth.</subject> <samples>SAMPLE 1 (LinkedIn post, recent): 'Three years ago we had 14 people and $1.2M ARR. Today: 280 people, $47M ARR. Here's the one thing nobody tells you about scaling from $1M to $50M. It is not about product-market fit. We had that. It is not about funding. We had that too. It is about the operational violence required to go from a team that can fit in one room to a team that literally cannot all meet at once. You have to break things you built on purpose. The thing you celebrated at Series A is the thing you destroy at Series B. If you are precious about what worked, you will fail at the next stage. I broke our entire sales comp structure 9 months ago. Our AE team hated it. Our revenue is up 38%. The version of Farah from 2 years ago would have fought to keep that structure. She was wrong.' SAMPLE 2 (email to leadership team, quoted in a Forbes profile): 'Team — I read the quarterly reviews. We are fine. Fine is not what I want. Here is what I want. I want us to ship the compliance module by EOQ. I want the onboarding time under 48 hours. I want our NPS above 62. I do not want excuses about resourcing. I do not want dependencies to become reasons. I will approve whatever resources you need if you tell me specifically what is blocked. Stop managing up to me. Start unblocking yourselves. I will see you Monday. Farah.' SAMPLE 3 (podcast interview transcript, excerpt): 'I think the hardest thing — and I still think about this — is that when you are a founder and a woman and not-white and not-from-the-valley, everyone wants you to have a thesis about being those things. And I do not. I mean, I have experiences. I have dealt with things. But my identity is not my business strategy. My business strategy is the business strategy. And sometimes I feel pressure to make my identity do work for the brand, and I resist that pressure. Hard. I am not the Pakistani woman founder. I am the founder, who is also Pakistani, who is also a woman.'</samples> <project-type>Business memoir — 'founder-led growth' angle, 80,000 words, probably for HarperBusiness or similar imprint.</project-type> <project-length>80,000 words</project-length> <voice-challenges>She's a fast talker with a lot of conviction. Prose-wise she's direct to the point of bluntness. She resists being 'personal' — the book needs some personal material (her journey, her family, her Pakistani-Canadian immigrant experience) but she's resistant. My challenge: how do I write personal scenes in HER voice when she doesn't naturally write personally?</voice-challenges> <subject-feedback>She said: 'Don't make me sound soft. I'm not soft. If a sentence makes me sound like someone who apologizes, cut it.' Also: 'I hate business-book clichés. No 'journey.' No 'authentic.' No 'passion.' These words are banned from the book.'</subject-feedback>
Voice Profile: Farah Nasser
A ghostwriter voice profile is a structured 7-signal analysis (lexicon, syntax, rhythm, rhetorical moves, register, fillers/tics, thematic obsessions) producing a reusable reference document that prevents voice drift across a 50,000+ word ghosted manuscript. Based on professional ghostwriting workflows, voice-match failure usually happens NOT from lexicon mismatch (easy to fix) but from syntax + rhythm drift (almost invisible but structurally diagnostic). Your subject Farah Nasser has one of the cleaner ghostwriting profiles I've seen — her voice across LinkedIn post, internal email, and podcast interview is remarkably consistent in syntax, register, and rhetorical structure. This profile produces the specific voice fingerprint + the checklist for maintaining it + the special challenge handling for personal scenes she resists writing.
Voice Summary
Farah Nasser's voice is: declarative, unhedged, rhythm-short, anti-apologetic, anti-cliché, thesis-forward. She writes the way she presumably talks — in short declarative sentences that stack into arguments through accumulation rather than elaboration. She does not use 'I think' or 'maybe.' She issues claims. Her voice is unusual in that her professional and personal registers are nearly identical — she speaks about business and about her identity with the same compressed, direct syntax.
Signal 1: LEXICON
Words she uses (repeatedly across samples):
- 'Fine' as critique ('we are fine. Fine is not what I want.')
- 'Specifically' as a demand for precision
- 'Violence' in business context ('operational violence')
- 'Wrong' applied to past self (self-correction framing)
- Specific numbers ('$1.2M,' '14 people,' '38%,' '48 hours') — quantitative always
Words she avoids (conspicuously absent):
- 'Journey' — explicitly banned
- 'Authentic' — explicitly banned
- 'Passion' — explicitly banned
- 'Think' as a hedge (she uses 'think' only for actual thought, not hedging — 'I think about this' yes; 'I think maybe' no)
- Softeners: 'just,' 'actually,' 'really,' 'sort of,' 'kind of'
- Business-book clichés generally — 'pivot,' 'journey,' 'grit,' 'hustle,' 'authentic leadership'
Signature phrases (appear across samples):
- 'Here is what I want.' / 'Here is the thing.'
- 'I do not want [X]. I want [Y].' (the do-not-want / want structure)
- 'I am not [X]. I am [Y].' (identity reframing structure)
- Specific numbers as punctuation
Technical vocabulary:
- Fluent in finance/SaaS metrics (ARR, NPS, AE comp, Series A/B)
- Uses these without explanation (assumes reader knowledge)
- For the book, translate to accessible when needed BUT preserve the habit of specificity
Signal 2: SYNTAX
Primary sentence construction:
- Subject + verb + object. Declarative. Almost no subordinate clauses.
- 'We had that.' 'I broke our entire sales comp structure 9 months ago.' 'My business strategy is the business strategy.'
Preferred structures:
- Short-short-long pattern: Two short sentences stating claims, followed by one longer sentence providing the texture. 'It is not about product-market fit. We had that. It is about the operational violence required to go from a team that can fit in one room to a team that literally cannot all meet at once.'
- Parallel denial + assertion: 'I do not want X. I do not want Y. I want Z.' (from Sample 2)
- Reframing syntax: 'I am not the X. I am the Y, who is also X.' (from Sample 3)
Avoids:
- Subordinate 'because' clauses embedded mid-sentence. She puts reasons in separate sentences.
- Parenthetical asides (hates them)
- Long compound sentences with multiple conjunctions
- Sentences beginning with 'However,' 'Furthermore,' 'Additionally' — transition-dependent
Contractions:
- She does NOT contract in written form. 'I am not' not 'I'm not.' 'Do not' not 'don't.' This is a very distinctive signature — most founders contract. She doesn't. Preserve.
Signal 3: RHYTHM
Sentence length profile across samples:
- Average: 11.4 words
- Shortest: 3 words ('We had that.')
- Longest: 33 words
- Standard deviation: ~7 words
- Ratio of sentences under 8 words: ~40%
Paragraph rhythm:
- Most paragraphs 4-7 sentences
- Usually open with a very short sentence (3-6 words)
- Middle sentences are medium (8-15 words)
- Often end with a short, punchy closer (4-8 words)
- Pattern: short → medium → medium → short. Bookend rhythm.
Distinctive rhythm moves:
- 3-beat stacks: 'I want X. I want Y. I want Z.' Three parallel short claims in succession.
- Short-sentence drumbeat: Multiple 3-5-word sentences in a row for emphasis. 'We had that.' 'She was wrong.'
- Self-correction closer: Paragraph ends with a short sentence revising what was just said. 'The version of Farah from 2 years ago would have fought to keep that structure. She was wrong.'
This rhythm is highly distinctive. Anyone trained to 'write well' is tempted to smooth this out with longer connective sentences. DO NOT DO THAT. Her rhythm IS her voice.
Signal 4: RHETORICAL MOVES
Signature patterns:
1. Setup + denial + claim: States what something 'is not about,' then what it 'is about.' Used in Sample 1 to frame the whole post.
2. Past-self correction: References 'the version of me from X ago' — uses this to show growth without the 'journey' cliché she hates.
3. Specific-numbers-as-proof: Backs every claim with a quantitative specific. 14 people, $1.2M, 38%, 48 hours, NPS 62.
4. Parallel structures: 'I do not want X. I do not want Y. I want Z.' — creates rhythm AND clarifies priorities.
5. Identity reframing: 'I am not the [categorical identity]. I am the [primary identity], who is also [categorical].' Reclaims self-definition.
6. Action framing: 'I broke X.' 'I read X.' 'I will approve X.' She centers her own agency. Rarely uses passive voice.
Signal 5: REGISTER
Primary register: Direct, declarative, professional.
Remarkably:
- Her LinkedIn post register = her internal email register = her podcast register. She does not code-switch significantly between public/private professional contexts.
- She is WARM through directness, not through softeners. 'Team —' opens the leadership email. That's as warm as she gets.
- She is PERSONAL through specificity, not through emotional disclosure. The podcast reveals identity-pressure WITHOUT revealing feelings about it. 'I resist that pressure. Hard.' = as emotional as she gets.
For the memoir: The book will NEED personal material. Her resistance is real. The move: write personal scenes in her register, NOT in typical-memoir register. She'll share facts + specific actions + past-self corrections. She will NOT share feelings-as-feelings. If a scene requires emotional content, FIND how she'd deliver it — via action or specific detail, not via 'I felt sad.'
Signal 6: FILLERS / TICS
Present in her voice:
- 'Here is...' opener ('Here is what I want,' 'Here is the thing')
- 'Specifically' as intensifier
- 'I will' future-declarative (rather than 'I'm going to')
- Short-sentence self-correction
- Numbers as periodic grounding
Absent (never use in ghosted prose):
- 'I feel like'
- 'I think'
- 'Maybe'
- 'In a way'
- 'To be honest' (she IS honest — doesn't need to announce it)
- 'Obviously'
- 'Look,' / 'So,' as sentence-starters (she's not folksy)
- Em-dash asides (she uses em-dashes sparingly, only for sharp pivots)
Signal 7: THEMATIC OBSESSIONS
What she returns to across samples:
1. Self-correction / past-self framing — shows up in all 3 samples. She positions growth as 'my past self was wrong about X.' This is one of her central voice moves.
2. Operational specificity — not 'we scaled' but 'we went from 14 people to 280 people.' She grounds abstractions in numbers constantly.
3. Resistance to identity-flattening — she does not want her business story subordinated to her identity story. This shows up explicitly in Sample 3 and will shape the whole book's treatment of her Pakistani-Canadian, female, non-valley-native identity.
4. Rejection of softening — 'Don't make me sound soft' is a meta-theme. Every paragraph should pass the 'would she apologize here?' test — and the answer must be no.
5. Compression as ethos — she does not elaborate unnecessarily. Short is the moral position. The book should be tight — probably under 80,000 words despite the contract stating 80,000. She will want it shorter.
The Gaps (What She Can't Do)
Farah's voice has specific gaps. Matching the gaps is essential.
1. Can't do long reflective passages. 2-3 sentences of reflection max. The business memoir format often wants 4-5 paragraph reflective openings. Do NOT write those in her voice.
2. Can't do traditional emotional disclosure. If a scene needs emotional weight, convey via action + detail, not feeling-language.
3. Can't do business-book-cliché scenes. No hero's-journey framing. No 'dark night of the soul.' She'll find those cringe.
4. Can't do folksiness / colloquial warmth. She's warm through directness, not through 'so here's the thing, folks.' Avoid.
5. Can't do hedged claims. Every claim in her voice is unhedged. If a claim needs hedging for accuracy, either find the unhedged version or restructure the sentence so the hedge is a fact, not a stance.
Special Challenge: Personal Scenes She Resists
She resists personal material. The book needs some. Here's how to write personal scenes in her voice:
Template: 'Action + specific detail + past-self correction.'
Example — childhood scene:
Wrong (typical memoir voice): 'Growing up as the daughter of Pakistani immigrants in suburban Toronto, I often felt caught between two worlds. My mother expected me to be dutiful; my teachers expected me to be exceptional.'
Right (Farah voice): 'My mother wanted me to be dutiful. My teachers wanted me to be exceptional. I was 14. I decided I was going to be exceptional, and the dutiful could wait. My mother was furious. She was also, partially, right. I understood that 20 years later.'
Short sentences. Specific age. Action verb ('I decided'). Past-self correction ('She was also, partially, right').
Run every personal scene through this template. If the draft reads like Joan Didion, you've missed her voice. If it reads like Farah, it's working.
Turing Test Setup
Before submitting chapters to Farah, run this test:
1. Write 3 paragraphs of ghosted prose.
2. Combine with 3 actual Farah paragraphs (from LinkedIn, email samples, etc.).
3. Interleave randomly.
4. Show to someone who knows her voice well (her COO? Her chief of staff?).
5. Ask them to pick which 3 are fake.
Target: They get it wrong at least 30% of the time. If they get all 3 right, your voice match is failing. If they get 2 of 3 wrong, your voice is landing.
Do this early in the project (first 3 chapters) and again midway (chapters 7-8). Voice drift is real and late-project.
Reusable Voice Profile Document
Print this out and keep it visible while writing:
The Farah Voice Checklist
Sentence-level:
- Average 11 words, range 3-30
- No contractions ('I am not,' not 'I'm not')
- No softeners: just, really, actually, maybe, kind of
- No em-dash asides (only em-dashes for sharp pivots)
- Short-medium-medium-short paragraph rhythm
- At least one 3-6 word sentence per paragraph
Word-level:
- BANNED: journey, authentic, passion, pivot, grit, hustle, authentic leadership, 'to be honest,' 'I feel like'
- REPEAT PHRASES ok: 'Here is what I want,' 'I do not want X. I want Y,' 'I am not X. I am Y.'
- Numbers: specific, quantitative, always
- Self-correction: 'The version of me from X ago was wrong about Y.'
Rhetorical:
- Setup + denial + claim
- Past-self correction (not journey)
- Specific numbers as evidence
- Action framing (I broke, I read, I will approve)
- Identity reframing when relevant (not X. Am Y, who is also X)
Register:
- Direct always
- Warm through directness, not softeners
- Personal through action + detail, not emotion-language
- Same register public, internal, private
Gaps to respect:
- No long reflective passages (max 3 sentences)
- No traditional emotional disclosure
- No hero's-journey framing
- No folksiness
- No hedged claims
Test:
- Would she say this?
- Is there a shorter version?
- Would she apologize here? (answer: no)
Key Takeaways
- 7 voice signals: lexicon, syntax, rhythm, rhetorical moves, register, fillers/tics, thematic obsessions. Syntax + rhythm are the hardest to fake and most diagnostic.
- Farah Nasser's voice is declarative, unhedged, rhythm-short, anti-apologetic, thesis-forward. Her professional and personal registers are nearly identical.
- Match the GAPS as carefully as the strengths. She can't do long reflective passages or emotional disclosure — your ghosted prose can't either.
- For personal scenes she resists writing: use 'action + specific detail + past-self correction' template. Never typical-memoir voice.
- Run Turing tests at chapters 3 and 7-8. Target: readers who know her get it wrong 30%+ of the time. This catches voice drift before she does.
Common use cases
- Professional ghostwriters starting a new book project
- Executive communications staff writing speeches/op-eds for a CEO
- Agency copywriters maintaining consistent brand voice across campaigns
- Political speechwriters capturing a candidate's voice
- Memoir ghostwriters (where voice-match is essential for authenticity)
- Collaborators writing 'as-told-to' books
- Thought-leader content agencies maintaining founder voice across blog posts
- Writers trying to parody or pastiche another writer's style (literary exercise)
- Translators capturing source-author voice in target language
Best AI model for this
Claude Opus 4 or Sonnet 4.5. Voice analysis requires deep linguistic attention + stylistic sensitivity. Top-tier reasoning essential.
Pro tips
- Get 3 samples minimum, 5 is better. Different genres (tweet, email, speech, longform) capture different voice facets. Same-genre-only samples produce incomplete profiles.
- The subject's TWEETS are often the best voice sample. Short-form writing strips pretense — what's left is pure voice. Long-form often shows edited voice, not native voice.
- Lexicon is the easiest signal (words they use, words they avoid) but the weakest as a match. Any writer can mimic lexicon. Real voice-match requires SYNTAX and RHYTHM — which are 10x harder to fake.
- Pay attention to what the subject CAN'T do. Every writer has gaps — can't land a joke, can't sustain a long sentence, can't do irony, can't be direct. Matching the GAPS is how you match the voice.
- The 'Turing test': write 3 paragraphs in the subject's voice, interleave with 3 real paragraphs, have someone who knows the subject pick which are fake. If they can't reliably pick, your voice match is working.
- Rhythm fingerprint: average sentence length, standard deviation, shortest/longest, rhythm of paragraph openings. Different writers have VERY different rhythm profiles — and rhythm is almost invisible to readers but structurally unique.
- Rhetorical moves are the most under-analyzed voice signal. Does the subject open with a claim, then support? Open with a story? Use hypotheticals? Use numbered lists? Does the SHAPE of their argument structure repeat?
- Don't match voice too perfectly on first drafts. Editors will sometimes read too-perfect mimicry as 'uncanny valley.' Match at 85%, leave 15% for the subject's natural correction instincts when they read.
Customization tips
- Run the voice analysis BEFORE drafting any chapters. Writing chapters first then trying to voice-match retroactively is backwards — you'll rewrite 3x. Voice profile first, draft in-voice from sentence one.
- Keep the subject's 3-5 samples physically open on a separate screen while drafting. Don't trust memory. Your voice-match will drift if you're only referencing your profile document.
- For book-length projects, re-analyze voice at the 30% mark. Subjects sometimes provide new interviews/drafts mid-project. Their voice may have shifted slightly. Update profile.
- If your subject is inconsistent across samples (different registers in different contexts), pick the PRIMARY context for your project and voice-match that one. Don't try to blend voices — the result is muddy.
- Deliver chapters with a voice-match confidence note: 'I've matched voice at ~88% per our profile. The personal scenes may feel uncomfortable to you initially — this is expected. Please correct where you'd phrase differently.' Managing expectations is the meta-voice-match.
Variants
Memoir Mode
For memoir ghostwriting. Emphasizes personal-voice signals — the subject's specific relationship to memory, emotion, disclosure.
Business/Thought-Leader Mode
For CEO content, executive communications, business books. Emphasizes register + claim-support structure + sector-specific lexicon.
Celebrity/Public-Figure Mode
For public figures with well-known voices. Adds public-voice vs. private-voice distinction + crisis-management language considerations.
Academic/Expert Mode
For academics or subject-matter experts. Handles the register challenge of making academic voice trade-book accessible without losing authority.
Frequently asked questions
How do I use the Ghostwriter Voice Matcher — Capture Anyone's Voice From 3 Samples prompt?
Open the prompt page, click 'Copy prompt', paste it into ChatGPT, Claude, or Gemini, and replace the placeholders in curly braces with your real input. The prompt is also launchable directly in each model with one click.
Which AI model works best with Ghostwriter Voice Matcher — Capture Anyone's Voice From 3 Samples?
Claude Opus 4 or Sonnet 4.5. Voice analysis requires deep linguistic attention + stylistic sensitivity. Top-tier reasoning essential.
Can I customize the Ghostwriter Voice Matcher — Capture Anyone's Voice From 3 Samples prompt for my use case?
Yes — every Promptolis Original is designed to be customized. Key levers: Get 3 samples minimum, 5 is better. Different genres (tweet, email, speech, longform) capture different voice facets. Same-genre-only samples produce incomplete profiles.; The subject's TWEETS are often the best voice sample. Short-form writing strips pretense — what's left is pure voice. Long-form often shows edited voice, not native voice.
Explore more Originals
Hand-crafted 2026-grade prompts that actually change how you work.
← All Promptolis Originals