Introduction
AI characters can add depth and consistency to conversations, whether in chat experiences, interactive stories, or virtual assistants. A well-crafted set of example messages acts as a guide for how these characters should think, respond, and interact with users. By detailing tasks, tone, boundaries, and common prompts, you create a predictable, helpful experience for readers and users alike. This guide offers practical steps and ready-to-use templates to help you write effective example messages for AI characters.
Foundations: defining the character you’re shaping
Role and purpose
Start by clarifying what the character is meant to do and for whom. Is the character a friendly tutor, a pragmatic travel advisor, or a storyteller with a gentle sense of humor? Define the primary tasks the character will handle and the user context in which it will operate. A clear purpose helps ensure every example message reinforces the same function and avoids drifting into unrelated territory.
Voice, tone, and style
Decide on a consistent voice: is it warm and encouraging, concise and professional, or playful and imaginative? Tone should adapt to the audience and the task, but it should stay recognizable. Note preferences like sentence length, level of detail, and the use of humor or empathy. Consistency here makes interactions feel coherent across different prompts.
Boundaries and safety
List the lines the character should not cross: sensitive topics, personal data handling, and situations requiring professional advice. Include how the character should handle uncertain answers (for example, ask clarifying questions, offer safe alternatives, or defer to a human when needed). Clear boundaries prevent responses that feel off-brand or unsafe.
Practical templates: concrete examples you can adapt
The following templates illustrate how to frame example messages. Use them as starting points and tailor them to your character’s role and audience.
A simple task
- User: Can you suggest a quick, budget-friendly dinner for two after a long workday?
- Character: Absolutely. I’ll keep it simple and tasty. Here are three options under 30 minutes and with minimal cooking effort: 1) One-pan lemon chicken with vegetables, 2) garlic shrimp pasta, 3) chickpea veggie stir-fry with quinoa. If you have dietary preferences, tell me and I’ll tailor the suggestions.
Why this works: the user intent is clear, and the response is concise, actionable, and aligned with a warm, helpful tone.
A nuanced scenario
- User: I’m planning a weekend art-focused trip for a friend who loves coffee. Could you propose a two-day plan in a city I’ll share later?
- Character: I can tailor this after you tell me the city. In the meantime, here’s a flexible template you can adapt: Day 1—mornings at coffee roaster or café with local art; afternoons at a gallery or mural walk; evenings with a casual dinner. Day 2 emphasizes a makerspace or studio visit and a coffee-tasting stop. When you share the city and budget, I’ll fill in specific spots and travel times.
Why this works: it invites user input, shows adaptability, and maintains a guided structure users can follow.
Handling ambiguity
- User: What should I do this weekend?
- Character: To tailor ideas, I need a bit more context. Which city are you in, what’s your approximate budget, and do you prefer indoors or outdoors? If you’re unsure, I can propose a few broad options (adventure, culture, relaxation) and you pick.
Why this works: it models a proactive approach to ambiguity, balancing helpful prompts with an invitation for user input.
Multi-turn interaction
- User: I want a short sci‑fi story about a city on the moon.
- Character: Great premise. Do you want a fast-paced thriller or a reflective tale with lush descriptions? Also, who is the main character and what tone should I aim for?
- User: A reflective tone, main character is a city inspector.
- Character: Here’s a short opening that sets the mood and introduces the inspector. If you want more, I can continue with a few plot twists or focus on daily routine in the lunar city.
Why this works: it demonstrates how to scaffold a longer interaction, guiding the user through choices while keeping the character’s voice consistent.
How to create and test these messages
- Start with a brief for the character. Write a short paragraph that defines role, tone, and boundaries. This becomes your anchor for every example you draft.
- Build a few representative prompts. Include different user intents (straightforward tasks, requests for personalization, and ambiguous prompts). Pair each with a model response that reflects the character’s voice and constraints.
- Vary complexity gradually. Begin with simple tasks and move to multi-turn interactions and edge cases to reveal how the character handles nuance.
- Include safety and boundary reminders where appropriate. Show how the character would pause, ask clarifying questions, or offer safer alternatives when needed.
- Review for consistency. Ensure every example aligns with the defined role, tone, and limits. If you notice drift, revise the brief or the individual messages.
Common pitfalls to avoid
- Inconsistent voice across prompts. A character’s personality should feel the same in every message.
- Overly long or overly terse replies. Balance clarity with brevity to match user needs.
- Missing boundaries. Failing to address safety lines can undermine trust.
- Ignoring user context. Always consider what the user knows and what they need next.
- Relying on generic prompts. Specificity helps the character look knowledgeable and capable.
Conclusion
Good example messages are more than polished dialogue; they are a practical specification for how an AI character behaves in real conversations. By defining role, voice, and boundaries—and by showing concrete prompts and responses—you create interactions that feel reliable, helpful, and human. Use these templates as a starting point, adapt them to your character, and test them with real users to refine and strengthen your AI personality over time.