← The Schramko Playbooks

How to Build Your Voice File

Build a document that captures how you communicate so AI and team can match it.

By James Schramko

A voice file is a document that captures how you communicate. It helps AI tools, writers, and team members produce content that sounds like you.

This playbook shows you how to build one.

Why You Need This

Without a voice file, every piece of content requires heavy editing. Writers guess at your style. AI tools default to generic patterns. You spend time fixing tone instead of approving work.

A voice file solves this. It becomes the reference point for anyone creating content on your behalf.

What Goes In a Voice File

A complete voice file has seven sections:

1. Core Voice Principles
2. Sentence Construction Rules
3. Word Choices
4. Structural Preferences
5. Non-Negotiables and Hard Stops
6. Proof Standards
7. Calibration Examples

Each section answers a specific question about how you communicate.

Section 1: Core Voice Principles

This answers: What is the overall feel of your communication?

Write 5-10 short statements that capture your default approach.

Examples:

• Direct entry. Skip warm-up sentences.
• State the point first. Explanation follows if needed.
• Calm authority without performance.
• Plain language over clever language.
• Action orientation in every response.

To find yours, ask:

• When I read my best emails back, what do they have in common?
• What do people say about how I communicate?
• What is the opposite of how I sound?

Section 2: Sentence Construction Rules

This answers: How do your sentences actually work?

Document specifics:

• Average sentence length (count words in 10 of your sentences)
• Maximum sentence length you are comfortable with
• Do you use fragments? When?
• Active or passive voice?
• How do you start sentences?
• What punctuation do you use or avoid?

Examples:

• Average sentence length: 8-15 words.
• One idea per sentence.
• Active voice default.
• No semicolons in conversational content.
• Fragments allowed when emphasis serves clarity.

Section 3: Word Choices

This answers: What words do you reach for? What words do you avoid?

Create three lists:

Words you frequently use: Pull these from your actual emails, posts, and messages. Look for patterns. Note specific words that appear often.

Phrases you use: Capture your verbal tics and go-to expressions. These are often invisible to you but obvious to others. Ask someone who works with you.

Words and phrases you avoid or reject: These matter more than what you like. List words that make you cringe when you see them in drafts. Include industry jargon you hate, overused phrases, and anything that feels off-brand.

Section 4: Structural Preferences

This answers: How do you organise ideas?

Document:

• How you sequence information (do you lead with the answer or build to it?)
• How you construct arguments
• How you deliver conclusions
• How you handle transitions
• Your relationship with lists, headers, and formatting

Examples:

• Lead with the answer or recommendation.
• Follow with reasoning if needed.
• Evidence or example comes after logic.
• Next steps at the end.
• No summary restating what was already said.

Section 5: Non-Negotiables and Hard Stops

This answers: What immediately breaks your voice when violated?

Standard Non-Negotiables

These are patterns that make content feel wrong. Things that trigger immediate rejection regardless of how good the rest is.

Examples:

• Opening with flattery or validation.
• Explaining reasoning before stating the point.
• Multiple questions stacked in one response.
• Therapy language or emotional processing framing.
• Hype adjectives or manufactured urgency.
• Apologetic or self-deprecating language.

To find yours:

• Review drafts you rejected. What specifically bothered you?
• Look at content from others in your industry that makes you cringe. What patterns do you see?

Hard Stops (Absolute Fails)

Hard stops are different from non-negotiables. A non-negotiable is a strong preference. A hard stop means regenerate immediately, no exceptions, no context where it is acceptable.

To qualify as a hard stop, the violation must be:

1. Binary (either present or not, no grey area)
2. Context-independent (never acceptable in any form)
3. Identifiable by a fixed string or character match

Examples of hard stops:

• Em dash character. Never appears in any output. No exceptions.
• Contrast reframe constructions ("It's not X, it's Y." / "The real issue isn't X, it's Y."). This is an AI fingerprint. It signals templated thinking. Regenerate immediately if it appears.
• Any forbidden phrase you have added to your graveyard list (see below).

Enforcement Tiers

Voice enforcement has three tiers:

Hard Stop: binary fail. Regenerate immediately. No exceptions. Examples: em dash, contrast reframe construction, graveyard phrases.

Warn: flag for review before delivery. The output may still be usable but needs editing. Examples: en dash in non-numeric contexts, passive voice in an action-oriented section, excessive sentence length.

Advisory: patterns to watch over time. Not a failure in isolation, but a drift signal if they appear repeatedly. Examples: sentence starters repeating within a short piece, hedging language increasing over time.

The Graveyard

The graveyard is a list of phrases and constructions that have been banned from your voice outputs. It grows over time as you review and reject drafts.

When you catch a phrase that consistently makes content feel wrong, add it to your graveyard. Once in the graveyard, it becomes a hard stop.

Format example:

• "dive deep into" - banned
• "at the end of the day" - banned
• "game-changer" - banned
• "leverage your unique value proposition" - banned
• "the brutal truth is" - banned

The graveyard is not a style guide. It is a failure archive. Entries stay in unless you explicitly remove them.

Section 6: Proof Standards

This answers: How do you handle claims and evidence?

Document your rules for:

• When you can make definitive claims
• When you use hedging language
• How you handle testimonials and case studies
• Your approach to statistics and numbers

Examples:

• All claims verifiable. All testimonials real. All numbers accurate.
• If unverified, use possibility language: can, could, often, tends to.
• Never fabricate testimonials, statistics, or case studies.

Section 7: Calibration Examples

This answers: What does off-voice actually look like?

Write 3-5 example sentences that are clearly wrong for your voice. These help writers and AI tools understand the boundaries.

Examples of off-voice sentences:

• "So, I've been diving deep into this space and honestly, I think there's an incredible opportunity here for you to really transform your approach!"
• "Great question! To be honest, most people struggle with this."
• "At the end of the day, the brutal truth is that you need to leverage your unique value proposition."

The worse these examples feel to you, the more useful they are.

How to Build Your Voice File

Step 1: Gather raw material. Collect 20-30 examples of your actual communication. Emails you sent. Posts you wrote. Messages you are proud of. Transcripts of you speaking.

Step 2: Look for patterns. Read through your examples and note what repeats. Sentence length. Word choices. How you start and end. What you never do.

Step 3: Fill in each section. Work through the seven sections above. Use your examples as evidence. Do not write what you think you should sound like. Write what you actually do.

Step 4: Add calibration examples. Write 3-5 sentences that would make you reject a draft immediately. These are often more useful than positive examples.

Step 5: Build your graveyard. Review past rejected drafts and list every phrase that triggered a rejection. Add these to the graveyard section of your Non-Negotiables.

Step 6: Test it. Give the voice file to someone else. Have them write a short piece using only the file as reference. See what they get right and wrong. Update the file based on what was missing.

How to Use Your Voice File

With AI tools: Paste the voice file into your instructions or system prompt. Reference it when asking for content. Use the calibration examples to show what to avoid. Use the graveyard as a checklist before any output is approved.

With writers: Share the file before any project begins. Review their first draft against the non-negotiables. Update the file when you give feedback they could not have known from reading it.

With your team: Anyone creating content on your behalf should have access. The file replaces endless rounds of "that does not sound like me" feedback.

Maintenance

Review your voice file every 6 months. Your communication evolves. Update it when you notice:

• New phrases you have adopted
• Old phrases you have dropped
• Rules that no longer apply
• Gaps that caused problems
• New hard stops earned through repeated failures

A voice file is a living document. Keep it current.

Quick Start Template

If you want to begin immediately, copy this structure and fill in each section:

VOICE FILE: [Your Name]

1. CORE VOICE PRINCIPLES

[5-10 short statements about your overall approach]

2. SENTENCE CONSTRUCTION RULES

[Specifics about length, structure, punctuation]

3. WORD CHOICES

Words I use often: [List]

Phrases I use: [List]

Words and phrases I avoid: [List]

4. STRUCTURAL PREFERENCES

[How you organise and sequence information]

5. NON-NEGOTIABLES AND HARD STOPS

Non-negotiables: [Things that make content feel wrong]

Hard stops: [Absolute fails, binary, no exceptions]

Graveyard: [Specific banned phrases]

Enforcement tiers: [Hard Stop / Warn / Advisory classification]

6. PROOF STANDARDS

[Your rules for claims and evidence]

7. CALIBRATION EXAMPLES

[3-5 sentences that are clearly wrong for your voice]

Last updated: May 2026