Artifacts — When Claude Builds Something You Can Use

The difference between Claude answering you and Claude handing you a finished thing.

Working draft — 2026-06-06

---

Why This Asset Exists

When you ask Claude for something small, it answers in the chat. But when you ask for something substantial — a document, a plan, a table, a chart, even a little app — burying it in a wall of conversation would be useless. So Claude opens an Artifact: a panel beside the chat where the actual thing lives.

Most people don’t realize two things about Artifacts: that you can iterate on them in place, and that you can export them. That’s what turns Claude from “a chatbot that answers” into “a tool that produces.”

---

The One-Sentence Version

When Claude makes something you’d want to use outside the chat, it opens it as an Artifact — a live document you refine in place and then copy, download, or share.

---

What Opens an Artifact

Artifacts appear when Claude generates substantial, standalone content — the kind of thing you’d want to keep, not just read once:

• A document — a memo, a plan, a one-pager, an email.
• A table or checklist.
• A chart or diagram.
• Code.
• A simple working app or web page.

The rule of thumb: if it’s something you’d use outside the chat, it’ll usually land in an Artifact.

---

The Key Behavior: It’s Live

This is the part people miss. An Artifact isn’t a one-shot output you have to regenerate from scratch. It’s live — you refine it by just asking.

“Make it shorter.” “Add a section about onboarding.” “Change the tone to friendly.” Each request updates the Artifact in place — it rewrites itself, right there in the panel, while your chat stays on the left. You iterate, version by version, until it’s right.

You’re not starting over each time. You’re editing a living document by talking to it.

---

Chat on the Left, the Thing on the Right

Think of your screen as two halves:

• The chat (left) is the conversation — your thinking, your back-and-forth.
• The Artifact (right) is the deliverable — the actual document you’re building.

Keeping them separate is the point: you can have a messy, exploratory conversation on the left while the clean, finished thing takes shape on the right.

---

Export & Share

When it’s ready, an Artifact is a real deliverable. You can copy it, download it, or share/publish it — drop it into a doc, send it to a colleague, ship it. It’s not a transcript you have to clean up; it’s the finished product.

---

A Note on Tokens

Artifacts are made of output — so yes, they cost tokens, and a long Artifact costs more than a short one. (See Models, Tokens & Usage.) Worth knowing if you’re iterating heavily on a big document.

---

Common Confusions

“Where did my answer go?” If Claude opened a panel, your answer is in it — the Artifact on the right, not buried in the chat.

“Can I edit it myself, or only by asking Claude?” You refine it by asking Claude — that’s the main way. Then export it and edit wherever you like.

“Is an Artifact saved?” It lives with the conversation that made it. To keep it for good, export it (copy/download) or share it.

---

Landmark Language

• If Claude makes something you’d use outside the chat, it lives in an Artifact.
• An Artifact is live — you refine it in place, you don’t start over.
• Chat on the left, the thing on the right.
• An Artifact is a deliverable, not a transcript — copy it, download it, share it.

Say these often. The goal is for someone to stop copy-pasting walls of chat and start treating Artifacts as the finished work product.