Setting the Right Tone So Your Chatbot Sounds Like Your Brand
How Do You Set the Right Tone for a Chatbot?
You set the right tone by defining formality level, a specific list of dos and don’ts, and a couple of sample phrases pulled from your own actual writing — then testing the result against real questions before it goes live. This is the practical, step-by-step follow-up to how AI chatbots learn your brand voice: an actual process to work through, not just the theory behind it.
Step 1: Pick a Formality Level Honestly
Most businesses land somewhere between “formal law firm” and “casual coffee shop,” and it’s worth being honest about where yours actually sits rather than defaulting to whatever sounds most professional on paper. A financial advisory firm probably wants measured, precise language with minimal slang. A skate shop probably wants something closer to how staff actually talk on the floor. Match the tone to how your business genuinely communicates elsewhere — your emails, your social captions, your team on a phone call — not to some abstract idea of what a business “should” sound like.
Step 2: Write Specific Dos and Don’ts
This is where most tone setups go wrong — people write something vague like “friendly and professional” and expect that to actually shape output meaningfully. It won’t. Specific instructions work far better:
- Do: Answer the direct question in the first sentence, then add context.
- Do: Use contractions — “we’ll,” “you’re,” “don’t.”
- Don’t: Say “unfortunately” or “I apologise for any inconvenience” — reframe negatives directly instead.
- Don’t: Use corporate filler like “leverage,” “solutions,” or “at this time.”
- Do: Keep responses under four sentences unless the question genuinely needs more detail.
Five or six specific rules like these do more work than a paragraph of vague adjectives.
Step 3: Give It Real Sample Phrasing
Pull two or three actual sentences from your own writing — an email you’d genuinely send, a line from your About page, how you’d answer a common question on the phone. These samples anchor the model to your actual voice far more effectively than a description of that voice, the same way showing someone an example teaches faster than explaining a style in the abstract.
Step 4: Use a One-Click Suggestion as a Starting Point
Rather than starting from nothing, a system that can analyse your existing site content and suggest a tone profile based on how you already write gives you something concrete to edit rather than invent. You’ll typically still adjust it — maybe it’s slightly too formal, maybe it missed a phrase you specifically avoid — but editing is faster and more accurate than writing from a blank page.
Step 5: Test It Against Real Questions
| Test Case | What to Check |
|---|---|
| An easy, common question | Does it sound like your team, not a generic assistant? |
| A question it can’t fully answer | Does it decline gracefully, in your voice, without over-apologising? |
| A slightly frustrated visitor | Does the tone hold up, or does it slip back to generic AI phrasing? |
| The same question asked two different ways | Is the tone consistent across both? |
A Common Mistake Worth Avoiding
Setting the tone once at launch and never revisiting it. Read a few real transcripts a month or two in. Tone can drift subtly as the underlying AI model updates, and you’ll often spot a phrase creeping in that doesn’t match — “I understand your concern,” say — that’s worth explicitly adding to the don’t list.
How This Works in Practice
The Outview AI Chatbot builds this entire process into a guided setup wizard — formality, dos and don’ts, and a one-click tone suggestion drawn from your own site content — so you’re live with a chatbot that actually sounds like your business within minutes, not days of trial and error.
FAQ
How many dos and don’ts is too many?
Somewhere around eight to twelve specific, concrete rules tends to work well. Beyond that, instructions start to conflict with each other or get diluted, and the model struggles to weigh them consistently.
Should the tone change across different pages or contexts?
Usually not dramatically — consistency is part of what makes a brand voice recognisable. Minor context shifts (more concise for quick product questions, slightly warmer for support issues) are fine, but the core voice should stay stable.
What if different team members have different opinions on the right tone?
Treat it like any brand voice decision — pick one direction, document the specific rules, and test the output together rather than debating adjectives abstractly. Concrete examples settle these disagreements faster than descriptions do.
Key Takeaways
- Formality level, specific dos-and-don’ts, and real sample phrasing are what actually shape tone.
- Vague instructions do almost nothing — specificity is what makes the difference.
- A one-click suggestion from your own content saves time versus writing a profile from scratch.
- Test tone against easy questions, hard questions, and repeated phrasings before launch.
Try the guided tone setup in the Outview AI Chatbot — live in minutes.