Overview
How might we design dynamic follow-ups so people can trust and act on AI output?
When this pattern fits
- Ideal for search engines, conversational AI tools, and discovery applications where suggesting follow-up questions guides user exploration and improves engagement.
When to skip or lighten it
- Transactional exchanges where the task is complete and further prompting would feel like padding.
- Highly personal or creative threads where generic suggested questions break the user's own train of thought.
- Answers already ending in an explicit question or call to action that a chip would duplicate.
States
State model coming soon
Key UX elements
Key UX elements coming soon
Anti-patterns to avoid
Suggestions that just rephrase the previous question instead of extending it.
The same generic chips regardless of what the answer actually covered.
So many chips that they crowd out the actual response or next input field.
Chips that silently send a different query than their visible label implies.
How products use it
| Product | Implementation |
|---|---|
| Perplexity | Related questions appear below each answer, generated from the answer's content. |
| Bing Copilot | Suggested follow-up prompts sit under the response as tappable chips. |
| Google AI Overviews | "People also ask"-style follow-ups extend the search session contextually. |
| ChatGPT | Contextual suggested replies occasionally appear after certain response types. |
Implementation
Copy this prompt to generate a production-ready implementation in Cursor, Claude Code, Lovable, or any AI coding agent.
Generate a production-ready implementation of the "Dynamic Follow-ups" AI interface design pattern.
Pattern Definition:Frequently asked questions
What are dynamic follow-up suggestions in AI chat?
Dynamic follow-ups are clickable chips shown after an AI response that suggest logical next questions based on what was just answered, letting users continue exploring a topic with one tap instead of typing a new prompt.
How are follow-up chips different from prompt starters?
Prompt starters appear on an empty screen to help users begin a conversation. Follow-up chips appear after a response and are generated from that specific answer, so they extend an existing thread rather than starting one.
Do follow-up suggestions need to be personalized per answer?
Yes, generic suggestions that ignore the answer's content quickly feel like filler and get ignored. The suggestions should read as a natural next question a curious reader of that exact answer would ask.