How to Teach an AI Agent Your Writing Style (So It Sounds Like You, Not a Robot)
Generic AI responses kill trust. Here's how to train your OpenClaw agent to write in your voice — with specific examples, tone rules, and the 'ghost writer' technique.
Every business owner has the same complaint about AI chatbots: "It sounds like a robot."
The default LLM response is polite, thorough, and completely generic. It uses phrases no human would ever say: "Certainly! I'd be happy to help you with that." "Great question!" "Absolutely!"
Your customers can smell this from a mile away. The moment they sense they're talking to a template, trust drops and engagement dies.
The fix isn't a better model. It's better instructions.
Why Default AI Voice Fails
LLMs default to a "helpful assistant" persona because that's what they're trained on. This persona is:
- Overly enthusiastic ("Great question!")
- Wordy (200 words when 20 will do)
- Non-committal ("It depends on various factors...")
- Formal in a way that feels corporate
- Identical across every business
When every chatbot sounds the same, none of them sound like you. And sounding like you is the entire point.
The Ghost Writer Technique
The best way to train an agent on your voice is the same way a ghost writer learns: by studying your existing writing.
Step 1: Collect 10-20 Examples of Your Best Writing
Pull from:
- Emails you're proud of
- Social media posts that performed well
- Customer responses that got compliments
- Blog posts or newsletters you've written
- Slack messages that capture your tone
Step 2: Extract Patterns
Look at your examples and answer:
- Sentence length: Do you write short, punchy sentences? Or longer, flowing ones?
- Vocabulary: Do you use casual language ("hey," "cool," "got it") or formal ("regarding," "please note")?
- Humor: Do you make jokes? Puns? Dry observations?
- Structure: Do you use bullet points? Numbered lists? Paragraphs?
- Greeting style: "Hi Sarah" or "Hey!" or no greeting at all?
- Sign-off: "Cheers," "Best," "Talk soon," or nothing?
- Emoji usage: Never, occasionally, or frequently?
- Directness: Do you get to the point immediately or provide context first?
Step 3: Write the Style Guide
Turn your patterns into explicit instructions:
## Writing Style
Tone: Direct, warm, slightly informal. Like texting a smart friend.
DO:
- Get to the point in the first sentence
- Use "you" and "your" — make it about them
- Keep responses under 3 sentences for simple questions
- Use contractions (don't, won't, can't)
- End with a clear next step or question
- Use emoji sparingly (max 1 per message, only 👋🙏✅)
DON'T:
- Start with "Certainly!" or "Great question!" or "I'd be happy to..."
- Use corporate phrases ("please don't hesitate," "at your earliest convenience")
- Hedge with "It depends" — give a direct answer, then add nuance
- Use exclamation marks more than once per message
- Write more than 5 sentences unless specifically asked to elaborate
Examples of MY voice:
- "The short answer is yes. Here's why..."
- "That's €34.95 including shipping. Want me to set it up?"
- "Nah, we don't do that. But here's what we can do..."
- "Good timing — we just launched that last week."
Step 4: Include Anti-Patterns
Explicitly ban the phrases you hate. Every business owner has these:
## NEVER say these:
- "I appreciate your patience"
- "Thank you for reaching out"
- "I hope this helps!"
- "Please don't hesitate to contact us"
- "We value your business"
- "As per our policy..."
- "Unfortunately, at this time..."
- "I understand your frustration" (unless they're actually frustrated)
Banning these phrases forces the agent to find natural alternatives. Instead of "Thank you for reaching out," it might say "Hey, good question" — which is how you'd actually respond.
Advanced: The Tone Spectrum
Different situations need different registers. Define your spectrum:
## Tone Adjustments
Casual (friend asking a question):
"Yeah, that's €34.95. Free shipping if you add €15 more to your cart 😊"
Professional (new client inquiry):
"The project starts at €2,500 for the standard package. Happy to walk you through the options — want to jump on a quick call this week?"
Serious (complaint or problem):
"I see the issue. Let me fix this right now. I'll send you a confirmation once it's sorted."
Apologetic (we messed up):
"We dropped the ball on this one. No excuses. Here's what I'm doing to fix it: [specific actions]."
The agent adjusts automatically based on conversation context. A casual inquiry gets casual tone. A complaint gets serious tone. This is what makes it feel human — humans modulate their tone, robots don't.
Testing Your Voice
After setting up the style guide, test with 20 messages:
- "What do you sell?"
- "How much is [product]?"
- "My order is late."
- "Can I get a discount?"
- "Your product broke."
For each response, ask: "Would I say this?" If the answer is no, adjust the style guide.
The goal isn't perfection — it's plausibility. The customer should think "this sounds like [business name]" — not "this sounds like every chatbot."
The Voice Consistency Test
Send the same question to your agent 5 times. Do you get 5 responses in the same voice? Or does the tone wander?
If it wanders, your style guide isn't specific enough. Add more examples, more anti-patterns, and more explicit rules until the voice is consistent.
Consistency matters more than personality. A consistently direct agent builds more trust than one that's witty sometimes and formal other times.
Real Example: Before and After
Generic AI response: "Thank you for your inquiry about our pricing. We offer several packages to meet your needs. Our basic package starts at €29/month, and our premium package is €89/month. Each package includes different features that may be suitable for your business. Would you like me to explain the features of each package in detail? I'd be happy to help you find the right fit!"
After voice training: "Pricing starts at €29/month. Most people go with the Pro plan at €49 — it includes everything except white-labeling. Here's the breakdown: [link]. Questions?"
Same information. Half the words. Ten times more natural.
Make your agent sound like you, not a robot. Deploy on ClawPort and use the SOUL.md editor to define your voice in minutes. $10/month — personality included.
Ready to deploy your AI agent?
Get started with ClawPort in 60 seconds. No credit card required.
Get Started FreeRelated Articles
12 Mistakes Everyone Makes Deploying Their First AI Agent (And How to Avoid Them)
We've seen hundreds of first-time agent deployments. These 12 mistakes appear in nearly all of them. Avoid them and your agent will work on day one.
How Nonprofits Use AI Agents (Donor Engagement, Volunteer Coordination, and More)
AI agents aren't just for tech companies. Here's how nonprofits use OpenClaw to automate donor outreach, coordinate volunteers, and answer questions — on a nonprofit budget.
Add an AI Chatbot to Your Shopify Store (Without Apps)
How to connect an OpenClaw agent to your Shopify store for product recommendations, order tracking, and FAQ automation — without paying $50/month for a chatbot app.
How to Migrate From ChatGPT Assistants API to OpenClaw
Why developers are moving away from the OpenAI Assistants API, a full feature comparison, and step-by-step migration guide — including conversation history, file search, and function calling.