When Should AI Send a Greeting? Timing, Context, and Controls That Make It Feel Warm

What makes an AI greeting feel warm?

A greeting is a small invitation to begin an ordinary activity: continue a writing exercise, explore a creative prompt, or pick up a personal project. Its warmth comes less from cheerful wording than from fit. Does the moment suit the user? Is the suggestion relevant? Can the person choose whether to respond?

Research on proactive conversational agents describes initiative as a system starting an interaction rather than waiting for a user request. The authors argue that initiative can be perceived as intrusive without thoughtful design, and identify adapting timing and pacing to a person’s context and respecting their boundaries as important design dimensions. That offers a useful way to assess a greeting: look at the timing and the respect for choice, not just the friendly tone. “Towards Human-centered Proactive Conversational Agents,” arXiv

For example, a user who chose a brief afternoon creative prompt might appreciate a single invitation during that window. A greeting that arrives while they are already using the tool could instead feel like an interruption. These are illustrative situations, not claims about what every user prefers: the same moment may suit one person and not another. Letting people define their own preference is more reliable than treating a guessed routine as permission.

Which timing choices put the user in charge?

Start with timing the user explicitly selects: a chosen time, a chosen day, or a reminder tied to an activity they opted into. A broad preference such as “weekday afternoons” can be more practical than a rigid daily schedule, but the important point is that the person sets it and can revise it.

A greeting should also have a sensible stopping point. A single invitation during a selected window is easier to understand than a stream of follow-ups. If the person skips it, the system can let the moment pass; if they want another invitation later, they can choose that. This is an editorial recommendation drawn from platform guidance on making interruptions controllable and research that emphasizes adaptation and user preferences—not a universal frequency proven by the cited sources.

The W3C’s supplemental guidance on limiting interruptions recommends an easy way to control interruptions and reminders, including choosing which kinds appear and when. It also describes pausing them so people can return later. Applied to ordinary AI greetings, this suggests a practical default: make a greeting deferable, and let the user choose the next suitable time rather than assuming that silence means “try again soon.” W3C, “Limit Interruptions”

What context makes an invitation useful?

A useful greeting gives a concise, understandable reason for appearing. It might refer to a time window the person selected or offer a creative activity they asked to receive. The reader should be able to see the connection without having to infer what the system knows or why it has chosen that moment.

Keep the invitation proportionate to that reason. “Your afternoon sketch prompt is ready” is a clear invitation if the user chose an afternoon sketch prompt. It identifies the activity and leaves the next step open. If the system has no user-selected activity or timing to refer to, a generic greeting may be better left inside the app, where the person can choose to start it.

Context should not be used to make a greeting sound more personal than the underlying choice warrants. A good test is whether the user could explain why they received it in one sentence: “I asked for a prompt around this time.” If the explanation instead depends on hidden assumptions, the greeting may feel surprising even when the words are pleasant. This is a design inference from the emphasis on context, adaptivity, and civility in proactive-agent research; it is not a feature guarantee about any particular AI service. “Towards Human-centered Proactive Conversational Agents,” arXiv

Which controls should be easy to find?

Before enabling greetings, look for controls with plain names and direct effects. Useful choices include turning greetings on or off, selecting the types of invitation, choosing a time window, setting a frequency, and pausing them for a while. A visible “not now” action can handle a one-time skip; a readily accessible settings path should handle a lasting preference change.

Platform documentation makes clear that notification controls belong to the person using the device. Apple says apps should explain why they want notification permission in context, and that people can change notification settings at any time. Android provides notification channels that let people adjust how different notification types appear; users retain control over channel behavior after it is created. Apple Developer, “Asking permission to use notifications” and Android Developers, “Create and manage notification channels”

These are platform mechanisms, not proof that a particular AI product offers a custom greeting schedule or separate greeting category. Check the service’s own settings before relying on a control, and check device-level notification settings if a greeting arrives as an alert. Android’s guidance describes per-channel user settings; on Apple devices, the notification authorization and available interaction types are also controlled through system settings. The exact options depend on the product and device.

A simple controls check is to answer three questions: Can I turn this kind of greeting off? Can I change when or how it arrives? Can I skip one without changing everything? If the answers are visible and understandable, the user has more practical choice. The W3C guidance likewise recommends that interruption controls be easy to use and that users can decide which types they want and when. W3C, “Limit Interruptions”

How can you decide whether to enable proactive greetings?

Use this quick decision aid before switching them on:

Check: Moment; A user-respecting answer: I selected when a greeting may arrive, or I can keep it inside the app.

Check: Reason; A user-respecting answer: I can tell which opted-in activity or preference it relates to.

Check: Control; A user-respecting answer: I can skip, pause, adjust, or disable it without hunting through settings.

This table is a practical synthesis, not a scored test. If one answer is unclear, start with the least interruptive option available—for example, keeping prompts in the app until you decide to open it. If you do enable notifications, check after a few arrivals whether the timing still fits your routine. Change the setting when it does not.

The evidence points toward restraint and choice. A 2020 IBM Research paper proposed issuing, suppressing, or combining notifications based partly on user preferences and schedules; it presents a framework, not a guarantee that any specific scheduling method works for every person. The later human-centered-agent paper similarly treats adaptation and respect for boundaries as design considerations. Together with Apple, Android, and W3C guidance on user controls, these sources support a modest rule: an ordinary greeting should be an invitation the user can shape, not an expectation they must meet. IBM Research, “A snooze-less user-aware notification system for proactive conversational agents”

A practical rule for choosing a greeting

Enable an AI greeting when its timing is yours to choose, its purpose is easy to understand, and its controls are easy to reach. A concise invitation linked to a selected creative activity can make starting feel simple. If you cannot tell why it arrived or how to change the pattern, keep greetings off or use an in-app option until you can. Warmth is easiest to recognize when the user remains free to welcome, postpone, or decline the moment.

Keep exploring this topic