Artifacts and canvases
When the assistant produces a thing it pops into a side panel.
When the assistant produces something self-contained — a document, code, a webpage — it pops out into a side panel. Claude calls them artifacts; ChatGPT calls it Canvas; Gemini calls it Canvas too. Same idea, different label.
What can I help with?
This is a faithful imitation. Click a suggestion or type your own message.
What counts
- Long write-ups, briefs, plans
- Single files of code
- Mini-apps that preview live (HTML/React)
- Diagrams (Mermaid flowcharts, sequence)
Casual replies stay in the conversation. Anything you'd want to keep moves to the panel, where you can edit, iterate, copy, or download.
Artifact vs inline visualization
A newer wrinkle: Claude (and ChatGPT's Canvas mode) can also render charts or small widgets inside the conversation instead of breaking out into the side panel. Same code, different placement.
- Artifact / Canvas: side panel. Persists. You can come back to it, publish a share link, or download.
- Inline visualization: renders below the message it belongs to. Lighter-weight, good for one-off charts and "show me the analytics" prompts.
Rule of thumb: a thing you'd want to keep belongs in an artifact; a thing you'd want to glance at belongs inline. Most apps let you promote inline → artifact when you change your mind.