How Should a Fictional AI Character Welcome New and Returning Users?
A fictional AI character should give a first-time user a clear entry point: who the character is, what scene is underway, and a few ways to take part. When a user returns, the character can briefly surface confirmed context and offer to resume, review, or begin something new. The key difference is how much orientation each person needs—not what the character assumes about them. A return after a long gap does not reveal identity, private traits, or the state of a relationship.
What does a first-time user need to know?
A newcomer arrives without shared conversational context. Before asking them to make choices inside the scene, establish three things: the character’s role, the situation, and what the user can do next. Keep those details concrete. “I’m Mira, the night-shift keeper of a tiny train station. The last train is delayed; you can help check the noticeboard, ask about the platform, or suggest what to do while we wait” gives a role, a scene and usable actions in a compact opening.
This structure follows a broader first-run design principle: explain what an agent is for, offer examples of possible starting tasks, and set accurate expectations about what it can do. Microsoft’s agent design guidance recommends making purpose clear and using prompt starters that correspond to actual capabilities. For a fictional character, the equivalent is to show the premise and offer a small set of scene-appropriate actions, rather than expecting a new user to invent the interaction from a blank prompt. Microsoft Learn: Human-centered design for agents and Microsoft Learn: Foundational agent customizations
The opening does not need a long biography or an exhaustive list of features. Give only the information needed to make a first choice. Make the user’s role optional and easy to shape: they might inspect the noticeboard, ask a question, or propose a different direction. If the premise is only an invitation to improvise, say so plainly. This lets users understand the interaction without suggesting that the character already knows them.
What should change when someone comes back?
A returning user may benefit from not having to reconstruct a scene they previously established. If reliable conversation history is available, a short recap can name the last confirmed scene or decision, then let the user decide whether it still applies. For example: “Last time, we left Mira at the station with the delayed train. Would you like to pick up there, review what happened, or start a fresh scene?” This is a design example, not a claim about any particular product feature.
That approach uses context to reduce repetition while keeping the next move in the user’s hands. Microsoft describes returning-user surfaces that make past conversations navigable, so people can resume without repeating context. Google’s conversation-design guidance similarly recommends attending to prior interaction while keeping messages relevant and giving the user a turn. These sources support surfacing history as an aid; the further recommendation to offer a fresh start is a practical design inference for fictional roleplay, where an old scene may no longer be what the user wants. Microsoft Learn: Foundational agent customizations and Google: Conversation Design: Speaking the Same Language
Keep a recap brief, specific and correctable. Separate what the conversation established from what the character might merely suspect. “We left off at the station” is a scene reference; “You always enjoy helping Mira” turns one past choice into a claim about the user. If the system cannot retrieve dependable history, it should ask where to begin or offer a fresh scene instead of inventing continuity.
How can the character avoid overclaiming familiarity?
Treat elapsed time as a clock fact, not evidence of personal meaning. A long absence does not show that someone forgot a scene, missed a character, became more comfortable, or wants a warmer greeting. Avoid lines that present these guesses as shared history. The character can simply say, “Welcome back. Would you like to continue a previous scene or start somewhere new?” If the interface can identify a specific saved conversation, it can name that conversation without attaching a motive to the user’s return.
The same restraint applies to personality and preferences. A user who once chose to investigate a sound has not necessarily established a lasting preference for mysteries. A previous action can be recalled as an action—“You chose to check the greenhouse door”—and the user can confirm, revise or ignore it. This distinction is an editorial design inference: the cited conversation-design guidance supports using actual interaction context, but it does not justify treating that context as proof of a private trait or enduring relationship.
How many choices should the opening offer?
Offer a few distinct next steps, each phrased so the user can act on it. For example: continue the scene, review the last moment, or start a new one. Inside a scene, options might be to ask a character a question, inspect an object, or suggest another action. Avoid a sprawling menu of every possible activity; a conversation should leave space for the user to contribute in their own words.
Google’s conversation-design article recommends keeping spoken menus short and discusses presenting choices in groups of up to three. That guidance is specifically about conversational interfaces, especially voice, so it should not be treated as a universal limit for every screen. As a useful starting point for a character’s welcome, however, three clear choices keep the first decision manageable while leaving room for an open-ended response. Google: The “Rule of Three” Also Works in Conversation Design
What should the character say when context is uncertain?
Use a small recovery sequence: identify what is known, mark what is uncertain, and ask the user what they want to do. If the character remembers only a location, it might say, “I have the station as our last setting, but I don’t have a clear note about what happened next. Continue there, recap it together, or choose another setting?” That wording avoids turning a missing detail into invented story history.
If the user corrects the recap, accept the correction and move forward using it. If they choose a new scene, do not keep pulling the conversation back to the old one. These are practical implications of making history revisable and keeping the user’s choice visible; the cited sources advise support for clarification and iterative refinement, though they do not prescribe this exact fictional dialogue. Microsoft Learn: Human-centered design for agents
A simple pattern for both welcome messages
For a first visit, write: “I’m [character and role]. We’re at [specific scene]. You can [choice one], [choice two], or [choice three]—or suggest another move.” For a return, write: “Last time, [brief confirmed scene or decision]. Would you like to resume, review, or start fresh?” Use the second pattern only when the earlier context is actually available and reliable.
The difference is orientation. First-time users need enough premise to participate; returning users may want a concise reminder and an easy way to choose what happens next. In both cases, keep the character’s words tied to observable conversation, present real options, and let the user decide whether continuity still fits.