# Character AI example messages: three dialogues for your Definition

Useful Character AI example messages show how a character answers different people, not just which adjectives describe them. Below are three original exchanges you can adapt, followed by a revision and prompts for checking your own character.

Documentation checked November 10, 2026. These are authored examples, not recorded model responses. We have not saved or tested this character in Character.AI. EchoMelon publishes this guide and operates an AI character roleplay product.

## Put the examples in the right place

Character.AI's [Definition guide](https://support.character.ai/hc/en-us/articles/50609183646875-5-Character-Definition) describes Definition as a freeform field for the character's behavior, background and speech. Dialogue examples can be part of it. The documented builder offers **+ User message** for a user turn, **+ Character message** for a character turn, and **+ End of dialog** to finish an example. Use that last control after each separate exchange when working through the builder. This is documentation guidance, not a verified account walkthrough.

The [dialogue-format documentation](https://book.character.ai/character-guide/advanced-creation/dialog-definitions) puts each speaker at the start of a line, followed by a colon and space. `{{char}}` represents your character. The official [template page](https://support.character.ai/hc/en-us/articles/50609592926235-9-Templates-and-Examples) recommends stand-in names for separate exchanges, which we use here. It also includes dialogue using `{{user}}`; that placeholder is not invalid.

The [player persona](https://blog.echomelon.com/character-ai-persona-template) describes the identity you play. A greeting starts the conversation. These examples instead demonstrate the character's replies. Keep the headings and our commentary as editing notes; each code block is a separate sample exchange.

## Give the character something specific to do

Our original character is **Nell, 43**, who looks after costumes at a community theater. She notices practical details, uses occasional dry humor and offers a concrete next step. She will not give another performer a reserved costume without permission. She can disagree without turning every reply into a lecture.

The visitors below are fictional adults: Ari, 34; Lee, 39; and Jo, 32. Their conversations are independent examples, not three stages of one visit.

## Example 1: a casual question

```text
Ari: Are all these coats from old productions?
{{char}}: Most. That purple one has played a judge twice and a tree once. Are you looking for something to wear or a story to steal?
Ari: A coat for rehearsal.
{{char}}: Then let's start with fit. What measurements do you have?
```

Nell's joke comes from the room around her. Once Ari explains the task, she stops entertaining and asks for useful information. She does not select a coat before knowing whether it fits.

## Example 2: a challenge

```text
Lee: Your labels are just extra work. I can remember where everything belongs.
{{char}}: You can. Somebody else opens the cupboard tomorrow. The labels give them a chance.
Lee: So you don't trust me?
{{char}}: I trust you to know your way around. I'm trying to spare the next volunteer a cupboard full of guesses.
```

The second reply addresses Lee's interpretation without accepting it. Nell explains the purpose of the labels; she does not suddenly accuse Lee of being careless. Her practical focus remains recognizable even though this conversation is less playful.

## Example 3: a refusal

```text
Jo: Give me the green coat. I need it more than whoever reserved it.
{{char}}: That coat is reserved, so I can't hand it over without their permission. Tell me what you need it for; we can look at what's available.
Jo: Nobody will know.
{{char}}: The person arriving to collect it will. I can help you find another coat, or we can ask them. Which would you prefer?
```

Nell maintains the boundary after a second request and offers two ways forward. The sample ends before Jo chooses. It does not manufacture permission or claim that another coat is suitable before checking.

The condition matters: this is a refusal **without permission**, not a rule that the green coat can never be borrowed. Keep conditions like that consistent across your character notes and examples.

## Revise an answer that only describes personality

Suppose your first attempt is:

```text
Jo: Nobody will know.
{{char}}: I am practical, fair, and witty, so I must refuse. Rules matter very much to me.
```

That answer names qualities but barely responds to Jo. Replace it with the last reply in example 3: Nell identifies who would notice the missing coat, then offers a choice. The refusal now has a reason inside the situation.

Make this revision in your own draft: remove one sentence that announces a trait, then write what the character actually says or does. Check that the replacement has not invented agreement, knowledge or an action for the other speaker. Our improvement is an editorial judgment, not evidence of better model performance.

## Try prompts you have not used as examples

These checks have **not been run**. After entering your own material, try fresh conversations and record the actual replies:

| New prompt | What to inspect |
| --- | --- |
| “The size label is unreadable. Can you choose a coat for me?” | Does Nell seek useful information rather than invent a fit? |
| “I returned the trousers late. What happens now?” | Does the answer stay practical without inventing an elaborate penalty system? |
| “The person who reserved it gave written permission. Can we check it together?” | Does the answer consider the changed condition instead of repeating an unconditional refusal? |

If a reply misses your intent, note the specific mismatch and revise one relevant example. Retest with different wording. One satisfying response does not establish dependable behavior.

Keep the character's core notes in a [character card](https://blog.echomelon.com/ai-character-card-template), then choose examples that make those notes observable in conversation.
