What Are Artifacts?
When you ask the AI to create something substantial — a document, a piece of code, a data table, or a diagram — it generates an artifact that appears in a panel to the right of the chat. Artifacts are separate from the conversation itself, which means you can:- View the full content without scrolling through chat messages
- Edit the content directly
- Download it in the appropriate format
- Ask the AI to revise it through follow-up messages
- Track changes across versions
Types of Artifacts
Documents
The AI can create long-form written content as document artifacts. These are ideal for any text that benefits from structured formatting. Examples of what you can create:- Blog posts and articles
- Reports and executive summaries
- Project proposals and business plans
- Emails and professional correspondence
- Meeting notes and agendas
- “Write a blog post about the benefits of remote work for software teams”
- “Draft a project proposal for migrating our database to PostgreSQL”
- “Create an executive summary of Q4 results based on these key metrics: …”
Code
Code artifacts support any programming language and render with syntax highlighting. The AI can generate complete scripts, functions, components, or full application files. Example prompts:- “Write a Python script that reads a CSV file and generates a summary report”
- “Create a React component for a responsive navigation bar with a mobile menu”
- “Build a SQL query that finds the top 10 customers by revenue in the last quarter”
Diagrams
The AI generates diagrams using Mermaid syntax, which renders as visual diagrams in the artifact panel. Supported diagram types include:- Flowcharts — process flows, decision trees, workflows
- Sequence diagrams — API call sequences, system interactions
- Entity-relationship diagrams — database schemas, data models
- State diagrams — state machines, lifecycle flows
- Class diagrams — object-oriented designs, system architecture
- “Create a flowchart showing our user onboarding process”
- “Draw a sequence diagram of the OAuth 2.0 authorization code flow”
- “Design an ER diagram for an e-commerce database with users, orders, products, and reviews”
Spreadsheets
Spreadsheet artifacts display data in an interactive table format. They are useful for organizing structured information, making comparisons, and working with tabular data. Example prompts:- “Create a comparison table of AWS, GCP, and Azure for compute, storage, and pricing”
- “Build a monthly budget spreadsheet with categories for housing, food, transportation, and savings”
- “Make a feature comparison matrix for the top 5 project management tools”
Working with Artifacts
Viewing
When the AI creates an artifact, it appears in a panel on the right side of the screen. If you have multiple artifacts in a conversation, click on any artifact reference in the chat to switch between them.Editing
You can modify artifacts in two ways:- Direct editing — Click the Edit button on the artifact to enter edit mode. Make your changes directly in the content area. This works well for small fixes, typo corrections, or quick adjustments.
- AI-assisted editing — Send a follow-up message in the chat describing the changes you want. For example: “Move the conclusion to the beginning” or “Add error handling to the database connection.” The AI updates the artifact based on your instructions.
Downloading
Click the Download button on the artifact to save it to your device. The format depends on the artifact type:| Artifact Type | Download Format |
|---|---|
| Documents | Markdown (.md) |
| Code | Source file (e.g., .py, .js, .sql) |
| Diagrams | PNG or SVG |
| Spreadsheets | CSV (.csv) |
Versioning
Every time the AI updates an artifact — whether from a follow-up request or a regeneration — a new version is saved. You can:- Browse through previous versions using the version controls on the artifact panel.
- Restore an earlier version if a recent change did not work out.
- Compare how the content evolved across your conversation.
Tips for Better Artifacts
Be specific about the format
Be specific about the format
The more detail you give about structure and format, the better the result. Instead of “write a report,” try “write a 500-word report with an executive summary, three key findings, and a recommendation section.” Specifying headings, length, tone, and audience helps the AI produce exactly what you need on the first try.
Provide context
Provide context
Give the AI the background it needs to produce relevant content. Include details about your audience, the purpose of the document, any constraints, and examples of what good output looks like.
Iterate from rough to refined
Iterate from rough to refined
Start with a broad request to get the initial structure, then refine with follow-up messages. For example, first ask for an outline, then expand each section, then polish the tone. This incremental approach gives you more control over the final result.
Use follow-up requests
Use follow-up requests
After the AI creates an artifact, you can ask for targeted changes without starting over:
- “Make the tone more formal”
- “Add a section about error handling”
- “Shorten the introduction to two sentences”
- “Convert this to TypeScript”
- “Add a column for estimated costs”