All terms
Agents & tools
Artifact
Also known as: Canvas (ChatGPT), generated artifact
Claude's term for generated content rendered as a separate, persistent object — code, a document, a chart, a webpage — instead of just inline chat text.
What it means
An artifact in Claude is any substantial generated output that gets its own panel in the UI: a working React component, a markdown document, an SVG diagram, an HTML page, a Python script. Claude detects when its output is "artifact-shaped" (long, self-contained, likely to be iterated on) and renders it in a side panel where you can preview, edit, copy, or download it — separately from the conversation.
The pattern is now industry-standard. ChatGPT calls its version Canvas. Cursor and Claude Code treat code edits as durable file changes. Cohere has Coral. The unifying idea: don't bury the deliverable in chat scrollback. Pull it out, give it persistence, let the user iterate on it without re-prompting from scratch.
Artifacts solve a real UX problem. Without them, generating a 200-line component means scrolling a chat to find the latest version, copy-pasting carefully, and losing track of revisions. With artifacts, the latest version is always pinned, you can ask "make the button blue" and it edits the existing artifact instead of regenerating, and you can preview rendered output (a webpage actually renders, a chart actually displays) without leaving the chat.
For builders, artifacts are also a useful primitive. Anthropic's Claude.ai Artifacts can run client-side code in a sandbox — you can generate a working calculator, a small game, or an interactive data viz that just works in the chat. ChatGPT's Canvas is similar for documents and code. They blur the line between chat assistant and lightweight IDE, which is increasingly where AI tools are heading.
Example
You ask Claude to "build me a tip calculator." Instead of dumping HTML in the chat, it opens an artifact panel with a working, rendered calculator you can use immediately and iterate on with follow-up messages like "add a split-by-N option."
Why it matters
Artifacts changed the shape of how people use chat AI for real work. They turn a conversation into a working session where the deliverable is first-class. If you're building an AI product that generates non-trivial outputs, you need an artifact-style surface — chat-only UX is a dated pattern in 2026.