What is the most effective way to define a writing voice for AI?

Voice is multi-dimensional. Specifying formality, energy, complexity, warmth, and authority individually gives much more precise control than a single adjective, which the AI can interpret many ways.

How do you prevent voice drift in long AI outputs?

Voice drift happens in long outputs as the AI reverts to its default. Repeating voice specs, generating sections separately with voice reminders, and providing voice reference examples all help maintain consistency.

Voice & Persona

How to make AI adopt and maintain a consistent voice, personality, and writing style.

7 min read

3 quiz questions

Without style guidance, AI writes in a neutral, slightly formal, somewhat bland voice that sounds like a well-meaning textbook. This "AI voice" is a statistical average of all the writing in its training data. It's clear and correct but lifeless. To get writing that actually sounds human, you need to actively direct the AI's voice.

Voice isn't one thing — it's a combination of multiple dimensions. Specifying each dimension individually gives you much more control than a single adjective like "professional."

Formality: Casual → Conversational → Professional → Formal → Academic
Energy: Calm → Measured → Enthusiastic → Urgent → Intense
Complexity: Simple (6th grade) → Accessible → Moderate → Technical → Expert
Warmth: Detached → Neutral → Friendly → Warm → Intimate
Authority: Humble → Collaborative → Confident → Authoritative → Commanding

Multi-Dimensional Voice

Defines voice across multiple dimensions for precise control.

Write [CONTENT TYPE] about [TOPIC] using this exact voice profile:

Formality: [Casual / Conversational / Professional / Formal / Academic]
Energy: [Calm / Measured / Enthusiastic / Urgent]
Complexity: [Simple / Accessible / Moderate / Technical / Expert]
Warmth: [Detached / Neutral / Friendly / Warm]
Authority: [Humble / Collaborative / Confident / Authoritative]

Additional voice notes:
- Sentence length: [short and punchy / medium / long and flowing]
- Favorite punctuation: [em dashes / semicolons / exclamation points / periods only]
- Personality trait: [witty / earnest / no-nonsense / empathetic]

Sometimes the fastest way to define a voice is to reference a known writer, publication, or brand. This works because the AI has read those sources and can mimic the patterns.

Examples of "Write like" references: - "Write in the style of The Economist — confident, data-driven, slightly dry wit" - "Write like a Paul Graham essay — conversational, clear, building to a non-obvious insight" - "Write like Apple's marketing copy — simple, short sentences, benefit-focused, aspirational" - "Write like a Reddit comment — casual, direct, uses slang, gets to the point fast"

The "Write like" technique works best as a starting point, combined with specific adjustments. "Write like Paul Graham, but more formal and with data to support each claim" gives you the best of both approaches.

The hardest part of AI voice control isn't getting the right voice once — it's maintaining it across a long piece or multiple outputs. The AI tends to drift back toward its default voice over long outputs.

Define the voice at the beginning of the prompt with specific characteristics
For long outputs, add "Maintain this voice throughout — do not drift to a neutral tone" as a constraint
Break long content into sections and generate each separately with the voice spec repeated
Use a "voice reference" — paste 2-3 paragraphs of existing content in the voice you want and say "Match this voice"

Prompt Templates

Voice Cloner

Clones a specific writing voice from examples.

Here are 3 examples of writing in the voice I want you to match:

Example 1:
"[PASTE PARAGRAPH 1]"

Example 2:
"[PASTE PARAGRAPH 2]"

Example 3:
"[PASTE PARAGRAPH 3]"

Analyze this writing voice. Then write [CONTENT TYPE] about [TOPIC] in the same voice. Match the sentence structure, vocabulary level, personality, and rhythm. Do not make it more formal or more casual than the examples.

Brand Voice Guide

Creates a reusable brand voice guide from examples, then applies it.

Based on these examples of our brand writing:

[PASTE 3-5 EXAMPLES]

Create a brand voice guide that includes:
1. Voice summary (2-3 sentences)
2. Tone dimensions (formality, energy, warmth, etc.)
3. Vocabulary: words we use vs. words we avoid
4. Sentence structure patterns
5. Three "we write like this / not like this" examples

Then use this guide to write [NEW CONTENT TYPE] about [TOPIC].

Test Your Knowledge

Knowledge Check

1 / 3

Why does AI default to a bland, neutral writing voice?

Key Takeaways

✓AI defaults to bland writing because it produces the statistical average of its training data
✓Define voice across dimensions: formality, energy, complexity, warmth, authority
✓The "Write like" technique references known writers or brands as a shortcut
✓Voice drift is real — repeat voice specs and use reference examples for long content
✓Combine "Write like" references with specific adjustments for the best results

Previous Lesson Next Lesson

Continue Learning

Audience Adaptation

Adapt AI output for different audiences — from experts to beginners, executives to children.

6 min

What Are LLMs

A plain-English explanation of large language models and why they behave the way they do.

7 min

Tokens & Context Windows

Learn about tokens, context windows, and why they determine what AI can and cannot do.

6 min