Designing the empty state of a chat product

By Imran Sheikh · 2026-11-12 · AI in practice · 4 min read

Key points

  • The empty chat screen is the product’s first impression.
  • Say what the product does and offer concrete starting points.
  • Show limits early, so trust starts on honest ground.

A chat product opens on a blank box. For a new user, that blank box is a question without an answer. What can I ask. What will happen if I do. Is this safe to use. The empty state has to answer all three in a few seconds.

Most teams treat it as filler. A greeting, a logo and a cursor. This is a missed chance. The empty state is the only screen every user sees, and it sets the terms for everything after.

The blank box problem

A text field can accept anything, and that freedom is the problem. People freeze when the options are unlimited. They type something vague, get a vague answer and decide the product is not for them.

The job of the empty state is to shrink that gap. It tells people what the product is good at and shows them how to begin. It lowers the cost of the first message to almost nothing.

Say what this is for

Start with one plain line about the purpose. Skip the slogan and write something a new user can act on. Ask about your sleep logs. Plan a trip with a budget. Draft a reply to a hard email. The line should make a person think of a real thing they want to do.

Match the line to the product’s true strengths. If it handles three things well, name those three. Promising everything sets up disappointment. A narrow, honest promise is easier to keep and easier to trust.

Test the line by reading it aloud to someone outside the team. If they can repeat back what the product does in their own words, the line works. If they ask what it means, rewrite it with simpler words.

Offer starting points that teach

Suggested prompts are the most useful part of a good empty state. They give people something to tap. They also teach by example. A person who taps a suggestion learns what a good request looks like.

  • Write suggestions in the voice a user would use, not in marketing language.
  • Cover different kinds of tasks, so people see the range.
  • Keep each suggestion short enough to read at a glance.
  • Let a tap fill the box so the person can edit before sending.
  • Rotate or personalize them once you know something about the person.

Avoid generic prompts that could belong to any product. Tell me a joke tells the user nothing about what makes your product worth opening. Use real tasks that your product does well, drawn from what your users actually ask.

Show the limits early

Honesty belongs on the first screen. If the product makes mistakes, say so in a quiet line. If it is not a replacement for a professional, say that. If it stores nothing or stores something, say which. Early honesty costs little and pays back every time the product gets something wrong.

Keep this short and calm. One sentence under the input is often enough. A wall of disclaimers frightens people and gets ignored. A single clear line gets read.

For companion apps, tone matters even more. We make Emerge Care, a health companion, and the first screen has to feel steady and kind. The person may be tired or worried. The empty state should feel like a calm room.

Design for the returning user

The first visit is only one case. A returning user does not need a tour. They need to get back to what they were doing. The empty state should change with history.

  • Show recent conversations so people can resume them.
  • Offer follow-ups to things they asked before.
  • Hide the introduction once it has done its job.
  • Keep the input in the same place every time.
  • Let power users skip suggestions entirely.

The best empty states shrink over time. They start with plenty of help and give way to the user’s own habits.

Check the small things. Does the keyboard open at the right moment, or does it hide the suggestions. Does the layout hold on a small phone. Does it work with larger text. Does a screen reader announce the purpose line before the input.

Look at the loading state between the first tap and the first reply. A short, calm indicator tells people the system heard them. A frozen screen tells them something broke. Write the first response with as much care as the empty state itself.

Also think about what happens offline, and on a failed request. People will hit those states early. A clear message with a retry button keeps the first experience from ending in confusion.

Try it with a stranger

Hand the screen to someone who has never seen the product. Say nothing. Watch what they do in the first ten seconds. Do they read the line. Do they tap a suggestion. Do they type, and what do they type.

Their first message is your best research. If it matches your suggestions, the empty state is working. If it is something you never expected, you have learned what people really want from you. Feed that back into the suggestions.

The first screen should make the first message easy.

Measure what you can. Track how many people send a first message, and how long they wait before they do. A long pause often means the screen did not help. Compare versions of the screen against each other and keep the one that gets more people talking.

Spend an afternoon on your empty state. Write the purpose line five ways. Pick suggestions from real use. Add one honest sentence about limits. Few changes in a chat product pay back as much.

Journal · Book a call · Home