AI message - Usage
AI messages display a single assistant response inside a conversational thread. We recommend using them for answers users need to read, review and act on.

- Content
- Actions
Options
- Content: Keep the response easy to scan by structuring it (see content guidelines).
- Actions: Add only message-level actions that users expect after reading, e.g. copy, rate response quality or regenerate. We recommend using subtle tertiary icon buttons so actions stay available without competing with the answer.
- Sources: If the response is grounded in files, web results or internal data, expose that provenance close to the message content. Only display if there are dedicated sources to show.
Behavior in context
- Responsiveness: AI messages use from 45 to 80% of the chat's container width, depending on the viewport width.
Dos and Don’ts
- Do show clear loading indicators while the assistant is generating responses (see wording guidelines)
- Do use the same actions for each AI message for consistency, but not more than 4 to avoid overloading users, e.g. copy, feedback, regenerate
- Do add thumbs up or down actions only if you are aligned with data protection guidelines